uml-alf agent based adaptive learning framework: a...
Post on 10-Jul-2020
21 Views
Preview:
TRANSCRIPT
UML-ALF AGENT BASED ADAPTIVE LEARNING FRAMEWORK: A CASE STUDY ON UML
A THESIS SUBMITTED TO THE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCES
OF MIDDLE EAST TECHNICAL UNIVERSITY
BY
EFE CEM KOCABAŞ
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR
THE DEGREE OF MASTER OF SCIENCE IN
COMPUTER ENGINEERING
JULY 2010
Approval of the thesis:
UML-ALF AGENT BASED ADAPTIVE LEARNING FRAMEWORK: A CASE STUDY ON UML
submitted by EFE CEM KOCABAŞ in partial fulfillment of the requirements for the degree of Master of Science in Computer Engineering Department, Middle East Technical University by, Prof. Dr. Canan Özgen _______________ Dean, Graduate School of Natural and Applied Sciences Prof. Dr. Adnan Yazıcı _______________ Head of Department, Computer Engineering Assoc. Prof. Dr. Veysi İşler _______________ Supervisor, Computer Engineering Dept., METU Assoc. Prof. Dr. Kürşat Çağıltay _______________ Co-Supervisor, Computer Education and Instructional Technology Dept., METU Examining Committee Members: Prof. Dr. Petek Aşkar _______________ Computer Education and Instructional Technology, Hacettepe University Assoc. Prof. Dr. Veysi İşler _______________ Computer Engineering Dept., METU Assoc. Prof. Halit Oğuztüzün _______________ Computer Engineering Dept., METU Dr. Ayşenur Birtürk _______________ Computer Engineering Dept., METU Asst. Prof. Dr. Saniye Tuğba Bulu _______________ Computer Education and Instructional Technology, METU
Date:
I hereby declare that all information in this document has been obtained and presented in accordance with academic rules and ethical conduct. I also declare that, as required by these rules and conduct, I have fully cited and referenced all material and results that are not original to this work.
Name, Last name : Efe Cem KOCABAŞ
Signature :
iii
ABSTRACT
UML-ALF AGENT BASED ADAPTIVE LEARNING FRAMEWORK: A CASE STUDY ON UML
Kocabaş, Efe Cem
M. Sc., Department of Computer Engineering Supervisor : Assoc. Prof. Dr. Veysi İşler Co-Supervisor : Assoc. Prof. Dr. Kürşat Çağıltay
July 2010, 95 pages
As the amount of accessible and shareable knowledge increases, it is figured out that
learning platforms offering the same context and learning path to all users can not meet the
demands of learners. This issue brings out the necessity of designing and developing
adaptive hypermedia systems. This study describes an agent-based adaptive learning
framework whose goal is to implement effective tutoring system with the help of Artificial
Intelligence (AI) techniques and cognitive didactic methods into Adaptive Educational
Hypermedia Systems (AEHS) in the domain of Unified Modeling Language (UML). There
are three main goals of this study. First goal is to explore how supportive agents affect
student’s learning achievement in distance learning. Second goal is to examine the
interaction between supportive agents and learners with the help of experiments in Human
Computer Interaction laboratories and system analysis. The effects of the methodology that
agents give misleading hints which are common mistakes of other learners are also
investigated. Last goal is to deliver effective feedback to students both from IAs and tutors.
iv
In order to assess that UML-ALF has accomplished its objectives, we followed an
experimental procedure. Experimental groups have taken the advantage of adaptive and
intelligent techniques of the UML-ALF and control groups have used the traditional learning
techniques. The results show that there is a positive correlation between variables practice
score and number of agent suggestion which means, as the participants benefit from
supportive agents, they get higher scores.
Keywords: Adaptive Learning, Intelligent Tutoring Systems, Intelligent Agents, UML,
Distance education.
v
ÖZ
UML-ALF ETMEN TABANLI UYARLANABİLİR ÖĞRENME SİSTEMİ: UML ÜZERİNE BİR ÇALIŞMA
Kocabaş, Efe Cem
Yüksek Lisans, Bilgisayar Mühendisliği Bölümü Tez Yöneticisi : Doç. Dr. Veysi İşler Ortak Tez Yöneticisi : Doç. Dr. Kürşat Çağıltay
Temmuz 2010, 95 sayfa
Erişilebilen ve paylaşılabilen bilginin miktarı arttıkça, aynı içerik ve eğitim yolunu
sunan eğitim platformlarının tüm kullanıcıların isteklerini karşılayamadığı anlaşılmıştır. Bu
durum, uyarlanabilir eğitsel hiper ortamların tasarlanmasının ve geliştirilmesinin
gerekliliğini ortaya çıkarmıştır. Bu çalışmada, Birleşik Modelleme Dili (UML) alanında
yapay zeka teknikleri ve kavramsal eğitim metodlarının yardımıyla etkili bir öğretim
sisteminin uygulamaya döküldüğü etmen tabanlı bir uyarlanabilir öğrenme sistemi
tanımlanmıştır. Bu araştırmanın üç ana hedefi bulunmaktadır. Birincisi, destekleyici
etmenlerin öğrencilerinin başarısını nasıl etkilediğini araştırmaktır. İkinci hedef, destekleyici
etmenler ve öğrenciler arasındaki etkileşimi İnsan Bilgisayar Etkileşimi Laboratuvarı ve
sistem analizleri yardımıyla incelemektir. Akıllı etmenlerin, diğer kullanıcıların ortak
hatalarından elde ettiği yanıltıcı ipuçlarını kullandığı öğretim tekniğini de bu kapsamda
araştırılmıştır. Son hedef ise, öğrencilere akıllı etmenler ve eğitmenler tarafından verimli geri
dönüşler yapabilmektir. UML-ALF çalışmasının hedefine başarıyla ulaştığını belirlemek için
vi
deneysel bir prosedürün takip edilmesine karar verilmiştir. Deneysel grup, sistemin
uyarlanabilir ve akıllı tekniklerinin tümünden faydalanırken, kontrol grubu geleneksel
öğrenim tekniklerini kullanmaktadır. Sonuçlar kullanıcı puanı ile akıllı etmen önerileri
arasında pozitif bir korelasyon olduğunu, yani katılımcı akıllı etmenlerden faydalandıkça
puanının arttığını göstermiştir.
Anahtar Kelimeler: Uyarlanabilir Öğrenme, Akıllı Öğrenme Sistemleri, Akıllı Etmenler,
UML, Uzaktan Eğitim.
vii
To my parents
and
to my beloved
viii
ACKNOWLEDGMENTS
Numerous individuals provided support, guidance, and encouragement during
my study as a graduate student, and it is only appropriate that I acknowledge their
valuable contributions.
First of all, I would like to extend my deepest gratitude to my thesis advisor
Assoc. Prof. Dr. Veysi İşler for all of his enduring support, tireless revision, and
encouragement. I would like to express sincere gratitude to my co-advisor Assoc. Prof. Dr.
Kürşat Çağıltay for his valuable suggestions and comments.
I would like to express gratitude to my examination committee members.
I would like to express thanks to Assoc. Prof. Dr. Kürşat Çağıltay and other
members of Human Computer Interaction Research and Application Laboratory of METU
Computer Center who provides support for the study.
I would like to thank Murat Miran Köksal and Didem Tufam Avdan for their
encouragement and support.
I would like to express gratitude to my colleagues and friends for their
encouragement and support. I would also like to thank to all participants who joined in
experiments of the study.
Lastly, I would like to express my love and deep gratitude to all members of my
family and my girlfriend, İlkay Kayserilioğlu, for their patience, encouragement and for their
steady trust in me.
ix
TABLE OF CONTENTS
ABSTRACT ............................................................................................................................ iv ÖZ ........................................................................................................................................... vi ACKNOWLEDGMENTS ...................................................................................................... ix TABLE OF CONTENTS ......................................................................................................... x LIST OF TABLES ................................................................................................................ xiii LIST OF FIGURES .............................................................................................................. xiv ABBREVIATIONS .............................................................................................................. xvi CHAPTER
1. INTRODUCTION ............................................................................................................ 1 1.1 Background of the Study ...................................................................................... 1
1.1.1 Adaptive Educational Hypermedia Systems (AEHS) ...................................... 1 1.1.2 Intelligent Tutoring Systems (ITS) .................................................................. 2 1.1.3 Learning Management System ......................................................................... 4
1.2 Purpose of the Study .................................................................................................. 5 1.3 Research Questions .................................................................................................... 6 1.4 Significance of the Study ........................................................................................... 6 1.5 Assumptions ............................................................................................................... 7 1.6 Limitations of the Study ............................................................................................. 7 1.7 Following chapters ..................................................................................................... 7
2. RELATED WORK .......................................................................................................... 8 2.1 Adaptive Learning Environments .............................................................................. 8 2.2 Intelligent Tutoring Systems .................................................................................... 10 2.3 Intelligent Adaptive Learning Environments ........................................................... 13
3. PROPOSED MODEL .................................................................................................... 17 3.1 User Profiles ............................................................................................................. 17 3.2 User Modeling ......................................................................................................... 17
3.2.1 User Model's User Profile: ................................................................................ 18 3.2.2 User Model's Cognition Profile: ....................................................................... 18
3.3 Agent Modeling ....................................................................................................... 18 3.4 Framework ............................................................................................................... 20
x
3.5 AICC ........................................................................................................................ 21 3.6 Adaptive and Intelligent Techniques ....................................................................... 22 3.7 Intelligent Agents ..................................................................................................... 24
3.7.1 Diagnostic Agent (DA): .................................................................................... 24 3.7.2 Language Agent (LA): ...................................................................................... 25 3.7.3 Feedback Mechanism (FM): ............................................................................. 25 3.7.4 Supportive Agents (SAs) .................................................................................. 27 3.7.5 Model Agent (MA) ........................................................................................... 27 3.7.6 Adaptive Agent (AA) ........................................................................................ 28
4. METHOD ....................................................................................................................... 29 4.1 Implementation & Case Study: UML ...................................................................... 29
4.1.1 Unified Modeling Language (UML) Domain ................................................... 29 4.1.2 Training Process of UML-ALF ......................................................................... 30 4.1.3 Building Resources ........................................................................................... 37
4.2 Definition of Propositions ........................................................................................ 39 4.3 Participants of the study ........................................................................................... 39 4.4 Data Collection Procedure ....................................................................................... 40 4.5 Data Analysis Methodology ..................................................................................... 42
5. RESULTS ...................................................................................................................... 44 5.1 Proposition 1 ............................................................................................................ 44 5.2 Proposition 2 ............................................................................................................ 47 5.3 Proposition 3 ............................................................................................................ 53 5.4 Survey Results ......................................................................................................... 53
6. DISCUSSION ................................................................................................................ 62 6.1 The Contribution of Supportive Agents ................................................................... 62 6.2 The Interaction between Supportive Agents and Learners ...................................... 63
6.2.1 HCI experiments ............................................................................................... 63 6.2.2 Suggesting Common Mistakes as Misleading Hints ......................................... 63 6.2.3 Tracking Learner’s Movements ........................................................................ 64
6.3 Effective Feedback ................................................................................................... 65 7. CONCLUSION AND FUTURE WORK ....................................................................... 666
REFERENCES .................................................................................................................... 688 APPENDICES A. SCENARIOS OF BUILDING RESOURCES ............................................................ 722 B. SURVEY FOR DETERMINING EXPERIMENT GROUPS .................................... 755 C. EVALUATION SURVEY .......................................................................................... 766 D. HCI LAB RESULTS .................................................................................................... 81
xi
E. APPROVAL FROM ETHICS COMMITTEE ............................................................ 855 F. TEST QUESTIONS ..................................................................................................... 866
xii
LIST OF TABLES
TABLES Table 3.1 Description Of The Task Environment .................................................................. 26
Table 4.1 Resource Structure ................................................................................................. 38
Table 4.2 Adaptive Results Of Pilot Test .............................................................................. 41
Table 4.3 Content Of The Implemented Course UML .......................................................... 42
Table 4.4 Snapshot Of The Learning Nodes Of Chapter 3 .................................................... 43
Table 5.1 Descriptive Statistics Of Practice Results .............................................................. 44
Table 5.2 Statistics For Correlation Between # Of Agent Suggestion And Practice Score ... 45
Table 5.3 Statistics For Correlation Between # Of Mistakes And Practice Score ................. 52
xiii
LIST OF FIGURES
FIGURES
Figure 2.1 A Generic Structure To Implement MCQ Manageable By Actors ...................... 14
Figure 3.1 User Modeling ...................................................................................................... 19
Figure 3.2 Agent Modeling .................................................................................................... 20
Figure 3.3 Framework Layer ................................................................................................. 21
Figure 3.4 Workflow of the Framework ................................................................................ 23
Figure 4.1 UML-ALF Home Page ......................................................................................... 31
Figure 4.2 UML-ALF Registration Page ............................................................................... 32
Figure 4.3 UML-ALF Learning Path ..................................................................................... 32
Figure 4.4 UML-ALF Pre-Test .............................................................................................. 33
Figure 4.5 UML-ALF Course Tree ........................................................................................ 34
Figure 4.6 First Page of Object Oriented Modeling Course .................................................. 35
Figure 4.7 UML-ALF Practice Screen ................................................................................... 36
Figure 4.8 UML-ALF User Tracing Flow Chart ................................................................... 37
Figure 4.9 Output of the Resource Structure Stated in Table 4.1 .......................................... 39
Figure 5.1 Scatter Diagram for Practice Score And Agent Suggestion ................................. 46
Figure 5.2 Time To First Fixation .......................................................................................... 47
Figure 5.3 Fixation Count ...................................................................................................... 48
Figure 5.4 Visit Count ............................................................................................................ 50
Figure 5.5 Total Fixation, Visit Count and Mouse Clicks for Agent Area ............................ 51
Figure 5.6 Scatter Diagram for Practice Score And Number Of Mistakes ............................ 51
Figure 5.7 Result of Question 1 in Evaluation Survey........................................................... 54
Figure 5.8 Result of Question 2 in Evaluation Survey........................................................... 54
Figure 5.9 Result of Question 3 in Evaluation Survey........................................................... 55
Figure 5.10 Result of Question 4 in Evaluation Survey ......................................................... 55
Figure 5.11 Result of Question 5 in Evaluation Survey ......................................................... 55
Figure 5.12 Result of Question 6 in Evaluation Survey ......................................................... 56
Figure 5.13 Result of Question 7 in Evaluation Survey ......................................................... 56
Figure 5.14 Result of Question 8 in Evaluation Survey ......................................................... 56
xiv
Figure 5.15 Result of Question 9 in Evaluation Survey ......................................................... 57
Figure 5.16 Result of Question 10 in Evaluation Survey ....................................................... 57
Figure 5.17 Result of Question 11 in Evaluation Survey ....................................................... 57
Figure 5.18 Result of Question 12 in Evaluation Survey ....................................................... 58
Figure 5.19 Result of Question 13 in Evaluation Survey ....................................................... 58
Figure 5.20 Result of Question 14 in Evaluation Survey ....................................................... 58
Figure 5.21 Result of Question 15 in Evaluation Survey ....................................................... 59
Figure 5.22 Result of Question 16 in Evaluation Survey ....................................................... 59
Figure 5.23 Result of Question 17 in Evaluation Survey ....................................................... 59
Figure 5.24 Result of Question 18 in Evaluation Survey ....................................................... 60
Figure 5.25 Result of Question 19 in Evaluation Survey ....................................................... 60
Figure 5.26 Result of Question 20 in Evaluation Survey ....................................................... 60
Figure 5.27 Result of Question 21 in Evaluation Survey ....................................................... 61
Figure 5.28 Result of Question 22 in Evaluation Survey ....................................................... 61
Figure 8.1 HCI Participant 1 Class Diagram Practice ............................................................ 81
Figure 8.2 HCI Participant 1 Use Case Diagram Practice ..................................................... 81
Figure 8.3 HCI Participant 1 Activity Diagram Practice ....................................................... 82
Figure 8.4 HCI Participant 2 Class Diagram Practice ............................................................ 82
Figure 8.5 HCI Participant 2 Use Case Diagram Practice ..................................................... 82
Figure 8.6 HCI Participant 2 Activity Diagram Practice ....................................................... 83
Figure 8.7 HCI Participant 3 Class Diagram Practice ............................................................ 83
Figure 8.8 HCI Participant 3 Use Case Diagram Practice ..................................................... 83
Figure 8.9 HCI Participant 3 Activity Diagram Practice ....................................................... 84
Figure 8.10 HCI Participant 4 Class Diagram Practice .......................................................... 84
Figure 8.11 HCI Participant 5 Class Diagram Practice .......................................................... 84
xv
ABBREVIATIONS
AA: Adaptive Agent
AEHS: Adaptive Educational Hypermedia Systems
AI: Artificial Intelligence
AICC: Aviation Industry CBT (Computer-Based Training) Committee
AOI: Area of Interest
ARP: Automated Role Players
BSc: Bachelor of Science
BTN: Behavior Transition Networks
CAI: Computer Aided Instruction
DA: Diagnostic Agent
FM: Feedback Mechanism
GPS: Global Positioning System
HACP: HTTP AICC Communication Protocol
HCI: Human Computer Interaction
IA: Intelligent Agent
IMS: Instructional Management Systems
ITS: Intelligent Tutoring Systems
LA: Language Agent
LMS: Learning Management System
MA: Model Agent
MCQ: Multiple Choice Questions
PEAS: Performance Measure - Environment - Actuators – Sensors
PSI: Personalized System of Instruction
SA: Supportive Agent
SCORM: Shareable Content Object Reference Model
SME: Subject-matter Experts
TAO: Tactical Action Officers
UML: Unified Modeling Language
UML-ALF: UML - Adaptive Learning Framework
XML: Extensible Markup Language
xvi
CHAPTER 1
INTRODUCTION
1.1 Background of the Study
With the development of network technology and disadvantages of traditional
education, distance education has started to take an important role for method of learning.
Providing an effective distance education system becomes a hot research topic in the last
decade. Motivation is also a significant factor for an effective learning. Learners should take
an active role in their learning so that they are kept motivated. As the amount of accessible
and shareable knowledge increases, it is figured out that learning platforms offering the same
context and learning path to all users can not meet the demands of learners (Brusilovsky,
2001). This issue brings out the necessity of designing and developing adaptive hypermedia
systems. The main principle of adaptive learning is that no single didactic strategy is good
for all learners (Papadimitriou, Grigoriadou & Gyftodimos, 2009). As a result, learners will
be well-concentrated and well-motivated while they are dealing with their learning goals
when educational procedures are adapted to their user model. Each learner may have
different learning styles which expose the characteristic differences. Therefore each learner
reacts different to various educational approaches and methods (Kolb, 1984). There are
many researches that analyze learning styles and preferences of the learners in order to adapt
educational approaches according to their preferences (as cited in Papadimitriou et al. 2009)
1.1.1 Adaptive Educational Hypermedia Systems (AEHS)
Adaptive Educational Hypermedia Systems (AEHS) are considered to be the
solution to the problems of traditional e-learning systems. In such traditional systems ‘one-
size-fit-all’ approach is used whereas in AEHS it is personalized based on user’s abilities,
characteristics and knowledge. AEHS also provide solution to the ‘lost in hyperspace’
syndrome, in which user get confused due to the complex interface of hypermedia. Several
1
adaptive and intelligent techniques have been applied to AEHS so far (Papadimitriou et al.,
2009; Brusilovsky & Peylo, 2003; Brusilovsky, 2000). Adaptive techniques can be listed as:
Adaptive Presentation: Adaptive Presentation adapts the learning material according to
user’s background, characteristics and knowledge of the user model.
Adaptive Navigation Support: The purpose of this method is to guide user in hyperspace due
to the user model features. The most common ways of adaptive navigation support are direct
guidance, link sorting, link annotation, link hiding, disabling, and removal (Brusilovsky,
2000)
Adaptive Collaboration Support: Mainly the goal of this technique is to match different
students in a group according to their characteristics. However, in this work the purpose is to
use IAs as co-workers and explore the effect of collaboration.
Intelligent techniques can be classified as follows:
Curriculum Sequencing: one of the oldest ITS technologies. Its purpose is to present user a
learning path whose learning material is carefully selected for the user.
Intelligent Analysis of Student Solutions: After user submits the answer a solution analyzer
decide whether the result is correct or not. Then, feedback mechanism takes the initiative and
returns the response.
Interactive Problem Solving Support: Without waiting for the final answer of the user
helping during each step of the problem.
Example-Based Problem Solving Support: Its purpose is to help users solving the problem
by leading them to the related and successful solutions which they experienced before.
1.1.2 Intelligent Tutoring Systems (ITS)
Intelligent Tutoring Systems derived from Artificial Intelligence (AI). At the same
time AI was dealing with the idea of creating machines that can think like humans. As
researchers came across to the problems that were difficult to solve, they realized that trying
to imitate human cognition with computers needless, since human beings do not think like
computers. Thus, researchers headed to another research area like expert systems which was
more productive.
Intelligent tutoring systems consist of four different components: the interface
module, the expert module, the student module and the tutor module. The interface module is
the module where the interaction of the learner with ITS is provided. Generally interface
module is formed by a graphical user interface and it is sometimes a simulation environment
2
for a specific domain. Expert module is an agent-based intelligent environment including
knowledge-based description that ITS is teaching. Student module includes the information
of student's knowledge level and individualized preferences. Tutor module tries to take place
of the human tutor in ITS. It supports learners by providing feedback and remedial
instruction. (Feng, Heffernan, N.T., Heffernan, C. & Mani, 2009)
In the 1960s, researchers created a model called computer-aided instruction (CAI)
(Uhr, 1969). This model is basically designed to examine the learner with a problem, receive
and record the learner's response. Thus, it determines the achievement of the learner.
However, this system only assumes that when it produces learning material, learner would
acquire information easily. But, it does not find a solution to the way how people learn. An
intelligent tutoring system (ITS) is an advanced version of the CAI model. ITS is any
computer-based learning system that teaches learning materials and supports students
without the contribution of human tutors. The main principle applied in ITS is the theory of
learning by doing.
All along the time, artificial intelligence was an important area for ITS research.
There are too many learning systems which were built around expert systems. However,
another field of artificial intelligence identified as agents and multi-agent systems has
affected ITS research recently. Most conferences now focus on pedagogical agents
(Mengelle, de Léan & Frasson 1998). There are many reasons for coming out of this new
field. Agents have advantages of modularity and flexibility with the attitude of computer
science. There are several works studying knowledge gaining and learning skills for agents
up to now such as (Huffman, 1994), (Tecuci & Hieb, 1996). Agents might be described as
intelligent collaborators helping learners in achieving specific goals. They can make proper
actions and response to the change of the environment without the participation of users.
Russell and Norvig (2003, p. 32) defined an agent as ‘anything that can be viewed as
perceiving its environment through sensors and acting upon that environment through
actuators’. Agents can be categorized as the way they perceive the environment, including
human agent, robotic agent and software agent. A human agent uses its sense organs such as
vision, hearing, smell, taste, and touch for sensors and hands, legs, mouth and other body
parts for actuators. A robotic agent might have cameras, thermometers and GPS as sensors
and various motors for actuators. For a software agent, any input such as knowledge
database, file structures, network packets, keystrokes or mouse clicks might be its sensors
whereas any output such as writing files, printing to the screen might be its actuators. In this
3
study, we focus on software agents.
The agent function for an agent maps every possible percept sequence to a possible
action that agent can take. The agent program implements the agent function for an artificial
agent. A proper design for an agent program is contingent upon the nature of the
environment. A performance measure describes the standards for a successful agent
behavior. It evaluates the performance of an agent in an environment. Clearly, there is no
such performance measure suitable for all agents. An agent tries to maximize the expected
value for this evaluation. For this reason, a performance measure must be selected carefully.
Agents are grouped into five classes based on their degree of perceived intelligence and
capability: (Russell & Norvig, 2003) simple reflex agents, model-based reflex agents, goal-
based agents, utility-based agents and learning agents. It is also accepted that all agents can
improve its performance by learning.
1.1.3 Learning Management System
Although there is no universally accepted concept of the information society, it is
believed that it is started somewhere between the 1970s. While the total amount of
information doubled on each century till 1850s, this period decreased to 5 years in 1970s.
From 1980s, renewal of the information decreased to 1 year. Thus, total produced
information in last 30 years is greater than previous 5000 years (Somyürek, 2009). In such a
time that information is so intense and changing rapidly, it become obvious that business and
academic world need adaptive learning environments. Most companies and universities teach
their employees and students with Learning Management Systems (LMSs).
LMS is a common term generally used for large systems that manage online learning
events for students, instructors, and administrators. These systems mostly provide e-learning
programs, specific contents, training modules, and communication tools for their
participants. The most known open source LMSs are Moodle and Sakai, commercial LMSs
are WebCT and Blackboard.
E-learning, which is becoming widespread day by day and is being laid as
infrastructure by most of the universities, has both pros and cons. It is a student-centered
environment. A student may study a subject until learning it. There is no time limit. Student
may communicate with instructors and other students with various communication tools to
understand the topic and benefit from the experience of someone all over the world.
Therefore the response time for reaching the desired resource is very little. Cost of education
is also decreased. It is independent from time and space. Course materials can be reorganized
4
due to daily circumstances. Student can try himself with several exams and tests. Rapid
feedback can provide high motivation. All e-learning activities of someone can be reported.
However, instructors should know how to provide an efficient e-learning. It is difficult for
them to transfer traditional course material into online world. The requirements of e-learning
facilities and technical education and support for students and instructors are the other issues
to be considered. There are different e-learning types. They can be classified as web based
instruction, synchronous and asynchronous instruction, virtual education, computer based
distance education, internet based/aided education and online education. However,
traditional hypermedia systems such as LMSs that provide same page contents and
navigation links are inadequate in meeting the requirements of learners interested in
personalized learning desires. Demand of each learner differentiates due to their
characteristics, different learning styles, different information processing styles and choice of
using different knowledge resources. Adaptive hypermedia systems create a user model for
each user considering their various features and offer personalized settings. A system must
provide some circumstances in order to be an adaptive one. They need to have a hypermedia
system, this system ought to contain user modeling and system has to be adapted to the user,
using this user model. Adaptive learning systems are also closely related with Intelligent
Tutoring Systems and Artificial Intelligence. A learning content is adapted to each user and a
learning path is dynamically guided with the help of intelligent tutoring systems. While
adapting the information and its presentation, adaptive learning system uses artificial
intelligence. Main goal of adaptive hypermedia systems is to provide individualization of the
system according to specific needs and preferences of each learner.
1.2 Purpose of the Study
This study describes the UML-ALF – UML Adaptive Learning Framework – whose
goal is to implement effective tutoring system with the help of AI techniques and cognitive
didactic methods into Adaptive Educational Hypermedia Systems (AEHS). There are three
main goals of this study. First goal is to explore how collaborative learning – work affect
student’s motivation in distance learning. In this part, system uses artificial intelligence (AI)
techniques to provide Intelligent Agents (IAs) which take supportive agents’ place
(Remolina, Ramachandran, Stottler & Davis, 2009). Second goal is to deliver effective
feedback to students both from IAs and tutors. This is one of the research problems in the
field of Intelligent Tutoring Systems (ITSs) (Fossati, Eugenio, Brown, Ohlsson, Cosejo &
5
Chen, 2009). Last goal is to present an effective AEHS which is adapted according to
characteristics of learners in the domain of Computer Science about UML. Among the
criticism about UML, especially interested in one of them – “Problems in learning and
adopting make UML problematic.” (Bell, 2004) – In this study, adaptive techniques and
intelligent techniques such as Adaptive Presentation, Adaptive Navigation Support,
Curriculum Sequencing, Intelligent Analysis of Student Solutions and Interactive Problem
Solving were used.
1.3 Research Questions
The following research questions are posed, in order to achieve the purpose of the
study:
1. Do supportive agents provide contribution to the learning achievement of learners
in distance education?
2. Are learners interacted with supportive agents?
2.1 Does the methodology that agents give misleading hints which are common
mistakes of other learners have positive effects on learning the subject.
2.2 Is the methodology which forces learners to renew their knowledge as they
make mistakes effective on their learning achievement?
3. Can intelligent agents and computer based tutors give effective feedback to the
learners which is close to the performance of human tutors?
1.4 Significance of the Study
As the amount of accessible and shareable knowledge increases, it is figured out that
learning platforms offering the same context and learning path to all users can not meet the
demands of learners (Brusilovsky, 2001). Also, the contribution that is provided by tutors in
classroom education does not exist in distance education. These issues bring out the
necessity of designing and developing adaptive hypermedia systems. It is expected that this
study determines how adaptive hypermedia systems and intelligent agents contribute on
distance education. Besides, this study gathers the information of which methodologies of
adaptive and intelligent techniques have positive effects on learning, which learning path the
participants follow, how they react to the learning system, with the help of analysing data of
intelligent agents, face-to-face trials and observations of experiments in Human Computer
Interaction Labs.
6
1.5 Assumptions
It is assumed that the participants of the study responded the questions properly,
accurately, and objectively. Moreover, it is accepted that language capability of the
participants are enough for learning materials.
1.6 Limitations of the Study
Adaptive part of this study was formed of unprofessional learning contents which
are not designed by the experts of UML domain. Thus, learning gain of participants is
limited to the success of these contents.
Because of the nature of distance education and time limitation, participants took
learning materials on their own computers. Consequently, the validity of results in learning
process is limited to the honesty of the participant behavior.
During data gathering process, HCI experiments were made with 5 participants due
to the time and scope limitations.
1.7 Following chapters
In Chapter 2, related works which have been done so far are being investigated.
Later on, in Chapter 3 proposed model of the UML-ALF is explained. User Modeling and
Supportive Agent modeling, Adaptive Educational Hypermedia Systems, adaptive engine,
feedback mechanism of the system and intelligent agents are represented in details. In
Chapter 4, a case study of this study - UML - is introduced. Also, implementation of the web
based hypermedia system is illustrated. In Chapter 5, evaluation of the system with detailed
analysis, graphs and charts are discussed. Lastly, conclusion of the thesis and future work is
stated in Chapter 6.
7
CHAPTER 2
RELATED WORK
This chapter is organized by using thematic review. The related works are grouped
and discussed in terms of the themes or topics they cover. The purpose of preferring thematic
review rather than chronological one is to demonstrate the types of topics which are
important to this research. There exist three groups including Adaptive Learning
Environments, Intelligent Tutoring Systems and Intelligent Adaptive Learning
Environments. All groups use intelligent agents in their systems. However, in first group just
adaptive techniques are used and similarly second group uses just intelligent techniques.
Third group is the hybrid of other two groups where both adaptive and intelligent techniques
are used.
2.1 Adaptive Learning Environments
Most LMSs have free-browsing learning mode. However, Hong, Chen, Chang &
Chen (2007) declare the idea that no such learning path is convenient for all learners and
argue that this mode decreases learner’s performance due to inappropriate course selection or
disorientation during learning period. Thus it is thought that deciding personalized syllabus
is a significant research issue for e-learning systems.
According to the work done by Hong et al. (2007), an appropriate learning path for
each individual is generated with the help of the result of pre-tests. In order to prove the
quality of this method, experimentation is made with some primary school students who are
dealing with the ‘Fraction’ subject. A learning path is generated according to the pre-tests.
Then, two sample groups are created. First group is trained by free-browsing method.
Second group used the learning path which is formed with the learning –path generation
algorithm. After the post-test results the achievement of first group is about %32 and second
group’s success is almost %48.
8
Due to successful results of this research, learning paths of users in our thesis are
determined with the results of pre-tests. Moreover, some questionnaires are used in order to
take more information about user, especially for adaptive learning in our study. Thus, more
effective learning paths are created.
Results can also decide the difficulty of the content and detail level. Experimental
results show that the proposed method can precisely provide personalized curriculum
sequencing based on difficulty parameter and concept continuity of successive courseware,
and moreover can accelerate learner’s learning effectiveness.
Cabukovski (2006) also presents the idea of offering different content views
according to the level of each user. He estimates the level of the student with an agent based
testing subsystem. This agent infrastructure is built on an existing distance education system
called MATHEIS which is a domain of mathematics and informatics. System supports XML,
SCORM and AICC standards. System has Supervisor Agent, User Agent, Level Agent and
Expert Agent. Briefly, all agents work for determining the level of user and presenting the
proper content to the learner. However, it is just one of the adaptive techniques - adaptive
presentation - of Adaptive Agent in our study.
Jin & Ming (2008) presents an intelligent agent system whose purpose is to find out
the interest of each user in different domains dynamically. They use Naive Bayesian
Classifier approach for determining user's interests. It provides recommendation for newly
created contents according to the inferences that has been previously made. System has three
kinds of intelligent agents: Personal Agent, Tutor Agent and Information Agent. Unlike
UML-ALF, this study works on multiple domains. Therefore, personal agent analyzes user
interests and can recommend interesting materials from different domains. Tutor agent, helps
users to find out proper materials when they ask for help and users notify information agent
about their demands and get response from information agent according to their pre-
specified demands. This study also aims to offer adaptive presentation for their users which
is just one of the duties of adaptive agent of our framework. However, since there are not
enough data, the accuracy of the system is stucked in 50%-60%.The advantage of this
research is analyzing and recommending various materials on multiple domains. But in our
study, a learning path is assigned for users adapting their user modeling and users are forced
to follow this path. The advantage of our framework is providing various adaptive and
intelligent techniques to the user during the whole process of learning.
Chen (2008) also uses intelligent agents for improving e-learning process by
applying an adaptive process. It suggests using portfolios which is a method to present self-
9
ability in different branches of industry for adjusting learners' education. In this research
intelligent agents are divided into two groups: Learning Assistants and Teaching Assistants.
These groups may stand for Study and Practice interfaces in our study. Agent based system
learning process starts with signing of the learner to the system. First User Agent collects the
user's portfolio then sends the necessary information to Learning Agent in order to present
adaptive learning material to the learner. Adaptive Agent which is responsible for the same
duty in our framework communicates with User Model in order to present adaptive material.
2.2 Intelligent Tutoring Systems
In order to develop computer programs, mostly software design methodologies are
used. Those methodologies are set of procedures which are followed during the development
process. It is considered that it is a creative process and cannot be reduced to standard
routines. However, it must have a design structure as well. Software design process mostly
consists of those steps: requirements, design, implementation, module test, integration test,
system test, acceptance and deployment.
Bowman, Lopez Jr., Donlon & Tecuci (2001) describe a learning environment that
software design methodologies are taught by intelligent agents. The user specifies the
requirements and the software development team, after gathering data and clarifying some
ambiguities, selects and uses a design methodology to develop the desired software product.
The process is capable of making an error. There is a difficult problems set and a group of
people who are highly qualified and are able to solve such problems. These definitions are
called as knowledge-based problems and subject-matter experts (SMEs).
Developing knowledge-based software agents that incorporate the expertise of the
SMEs is a widespread problem for AI field researchers (Bowman et al., 2001). Agents
consist of two main components: a knowledge base and an inference engine. The knowledge
base contains the data structures representing the entities from the expert's application
domain such as objects, relations between objects, classes of objects, laws, actions,
processes, and procedures. The inference engine consists of the programs that manipulate the
data structures in the knowledge base in order to solve problems in a way that is similar to
how the expert solves them. Knowledge base of Supportive Agents is held in Agent
Modeling and user knowledge base is defined with User Modeling in our framework.
In classical knowledge-based agent development process, there exists a
contradiction. A knowledge engineer is both part of solution and problem. In addition to this,
10
development of each new knowledge base or agent starts from the beginning, and it is a
difficult process. The goal is to enable SME that has no initial knowledge to develop a
knowledge base agent. This approach is called Disciple. It is based on developing learning
agents which are capable of using knowledge representation that makes easier knowledge
reuse and learning, and can be directly taught by an SME. Since each Supportive Agent and
user knowledge base starts from scratch there is no need for reusing the information in
UML-ALF domain.
There are several learning strategies such as learning from examples, learning from
explanations, and learning by analogy, in order to acquire the knowledge from the SME,
which are integrated by the learning and knowledge acquisition engine (Bowman et al.,
2001).
Knowledge base consists of two main components: object ontology and a set of
reduction and composition rules. Reduction rule is an if-then structure which expresses the
simpler problems that are reduced from the main problem. A composition rule is also an if-
then structure which combines the simpler solutions into a general solution of the problem
that is stated in reduction rule. Supportive Agents in UML-ALF also learn with reduction
and composition rules. Under this agent-building paradigm, knowledge engineers customize
the graphical user interface, help the SMEs to learn how to teach the Disciple agent and
make the reuse of ontological knowledge easier to support SME's creation of agents. This
paradigm eliminates much of the error generated by the many different people involved in
the typical framework for software development.
Remolina et al. (2009) introduce the study “Rehearsing Naval Tactical Situations
Using Simulated Teammates and an Automated Tutor” which is about developing a
simulation-based Intelligent Tutoring System for training Tactical Action Officers (TAOs).
TAO is the responsible officer in command center in navy ship of the entire watch team.
ITS’ role is to implement the same environment for the TAO to supervise. It uses Artificial
Intelligence algorithms to provide Automated Role Players (ARP) which are actually watch
standers that inform the TAO about the events in real environment.
Remolina et al. (2009) provide the communication between TAO trainees and ARPs
through a speech recognition unit, which understand TAO trainees’ commands in natural
language and transfer them to text and vice versa. ARPs are the equivalent of SAs in UML-
ALF. However, the communication between learner and SAs is provided by Language and
Diagnostic agents. The system is learn by doing tactical decision making ITS. Trainee
interacts with simulated teammates using a natural language interface. In order to simulate
11
the behavior of AI agents Behavior Transition Networks (BTNs) are used. It is accepted that
the most risky part of the project is the natural language interface, since it is based on
trainee’s speech output.
According to the grammar syntax of recognition system, user has to choose correct
commands. Before the advent of the system an instructor is needed for every two students,
but after this system now an instructor is needed for 42 students. Since BTNs are peculiar to
domain of this study, we preferred knowledge base decision support with simple reflex,
model and goal based agents in our study.
Similarly, Fossati et al. (2009) develop an ITS system which helps students to
understand the subject of Linked Lists easily. It is an important topic in Computer Science. It
is believed that most of the computer science students have difficulties in learning this
subject. According to the feedback of students, system is considered interesting, useful and
its performance is getting closer to the performance of human tutors.
Fossati et al. (2009) state that iList project is an interdisciplinary research whose
purpose is to understand the effective tutoring methods. It is believed that iList project has
reached some maturity that it can be used in computer science courses. According to the
results of this research, the importance of positive feedback is understood. iList is now
delivering mostly negative feedback to the mistakes of students. In order to generate more
feedback it is decided to monitor student actions more closely. UML-ALF delivers both
positive and negative feedbacks, with verbal and non-verbal responses. Feedback
Mechanism (FM) is responsible for this task and Diagnostic Agent help FM to produce more
effective feedbacks.
As well as Remolina’s approach, Faria, Silva, Vale & Marques (2009) also study
about developing a simulation-based Intelligent Tutoring System. Power Systems’
performance is needed to be effective. Control Center operators are responsible for dealing
with this task. Therefore, operators’ actions may end up with incident such as power failure.
The study is about developing an Intelligent Tutoring System (ITS) for training those
Control Center operators.
Many artificial intelligence algorithms are introduced during the development of this
system such as Multiagent Systems, Neural Networks, Constraint-based Modeling,
Intelligent Planning, Knowledge Representation, Expert Systems, User Modeling, and
Intelligent User Interfaces. It has two main tasks. One of them is Incident Analysis and the
other one is Diagnosis and Service Restoration. The system is used by Electrical Engineering
undergraduate students who will possibly be Control Center operator in the future. The
12
system is used for two purposes. For training BSc students which will probably will be hired
for job and for training already hired operators.
This environment is selected as one of the most important systems using AI
algorithms in Portugal (Faria et al. 2009). System has a connection with SPARSE, which is
an expert intelligent alarm processing system. It is also used for fault diagnosis training. It
provides model tracing technique to monitor operator’s reasoning. It either generates
scenarios from real cases or creates new scenarios for trainees. The difficulty level of
problems can be automatically assigned. With the help of Multiagent Systems paradigm,
modeling the interaction of several operators during system renewal is possible. Constraint-
based Modeling technique in restoration training can be used.
2.3 Intelligent Adaptive Learning Environments
Mengelle et al. (1998) use pedagogical intelligent agents inside Intelligent Tutoring
Systems (ITS). Agents' purpose is to help students to learn using pedagogical strategies. In
this manner, an agents and multi-agents system which is a field of artificial intelligence has
influenced this research.
Mengelle et al. (1998) focus on three main points: the actor (agent) paradigm,
communication techniques with the domain material and representation of this material.
Actors are defined as reactive, instructable, adaptive and cognitive agents. They have two
main features, react to the others and are able to learn. These agents are equivalent with
supportive agents (SAs) in our study.
They define an interpreted language to help building actors. Student is faced to two
types of agents: a teacher and a companion. When a student asks for help, companion dialogs
with the resource which includes complete answer and a hint. Companion should decide
whether to give a hint or complete answer. In our research learner only face with Supportive
Agents which is called companion in this study. After SAs suggest their opinions, a
scaffolding approach is used between tutor and learner in order to decide whether to give the
complete answer or a hint that leads learner to the answer.
Mengelle et al. (1998) state that the communication protocol between actors and
resources consists of four kinds of messages. Firstly, a given source is loaded with the basic
information such as its type. Then, resource informs the actors each time, when student
interacts with the resource. Resource may reply with several ways: reformulating question,
giving a correct hint, giving a misleading hint, giving complete answer. At last step, actors
13
decide what to do and choose one possible reaction due to their own expertise.
In order to be more effective, communication between the resource and supportive
agents is provided with the help of Language Agent and Diagnostic Agent in architecture of
our study. In addition to Mengelle, Dweck (2007), Izzo & Mazzone (2008) and Chialvo &
Bak (1999) also express that learning from mistakes is one of the learning methodologies.
Thus, in each practice some agents may give misleading hints which are common mistakes
of other learners in UML-ALF. But learners are informed of this situation in order to be
aware of the misleading hints. Also, tutor agent provides the correct hint using scaffolding
approach after SA’s suggestion.
While deciding the technique for building resources, the example of a classical tool
for knowledge assessment -Multiple Choice Questions (MCQ) - has been taken (Mengelle et
al., 1998). It allows straightforward diagnosis and quickly building of efficient resources.
The resource consists of four main parts: identification part, subject part, answer processing
part and a set of hints. Identification part is formed by an identification id and a list of related
knowledge. Subject part describes the question text and choices. An answer processing part
provides evaluating the knowledge of student and finally hints which may both help and
mislead student, are involved in resource.
Figure 2.1 A generic structure to implement MCQ manageable by actors
14
Similar resource structure is used in our framework. However, since domains of
these studies are different, techniques for building the resource are also different. Language
agent is responsible for building and interpreting the resource. The question designer links
each possible answer with a both mastered and not-mastered knowledge.
Mengelle et al. improve actor knowledge expertise by two tasks: diagnosis and
revision. The diagnosis task controls whether the behavior of the actor respect its goals. Each
actor has two main goals: its individual goal and the collective goal which is to help learner.
Each is goal is divided into sub goals. Diagnosis task checks every sub goal and find possible
problems. Then revision task aims to fix those problems to improve actor's knowledge.
All these processes actualize in our study almost in same way. Supportive agent
(SA) decides its answer according to its knowledge base while Diagnostic agent helps SA
communicating with Agent Model and resource. Also Diagnostic agent evaluates SA’s
performance and updates its knowledge level in agent model.
Papadimitriou et al. (2009) introduce a new adaptive educational system called
MATHEMA. The purpose of the system is to help high school students interactively by
eliminating their misunderstandings and learning difficulties. The domain of the research is
about electromagnetism in Physics.
Students may work in system both individually and collaboratively. Students may
follow an activity which is formed of experimentations, simulations, explorations, etc. The
experimental results show that following such an activity increase student’s achievement and
affect their motivation.
Adaptive Educational Hypermedia Systems (AEHS) can be a key solution to the
traditional online education systems. The problems of traditional systems are static content,
“lost in hypermedia” syndrome and ‘one-size-fits-all’ approach (Papadimitriou et al., 2009).
AEHS uses the ideas of Hypermedia and Intelligent Tutoring systems in order to
fulfill the requirement of producing such applications which can be adapted to the each
learner’s knowledge level, background, learning goal, style and cognitive preferences. In
order to introduce adaptation in web-based AEHS, several adaptive and intelligent
techniques have been applied. These are; Curriculum sequencing, Adaptive presentation,
Adaptive navigation support, Interactive problem solving support, Intelligent analysis of
student solutions, Example-based problem solving support, Adaptive collaboration
support—adaptive group formation and peer help. This study uses most of adaptive and
intelligent techniques that our framework uses, but not all of them. According to the
experiment results, none of the students answered all 12 questions correctly. After simulation
15
each of them increased their correct results. Then, with the help of collaboration process and
the dialog with the systems, at the end all of the students made 12 correct answers. An
experimental study with senior high school students showed that the MATHEMA helps the
students improve their performances. The evaluation techniques and results are so
impressive that evaluation methodology of our study is originated from this research.
16
CHAPTER 3
PROPOSED MODEL
In this chapter a model of adaptive educational system with supportive agent
infrastructure will be presented. This model offers an effective learning to the user which is
adapted according to individual skills and preferences of user. Also it supports learner during
his/her practice period with the help of agents using intelligent techniques. This model is
formed by a framework that includes knowledge base models (User Model, Agent Model),
data structures, agents (Adaptive Agent, Model Agent, Diagnostic Agent, Language Agent,
Feedback Mechanism, and Supportive Agents) and their communication. Following sections
analyzes the proposed model in details. A case study of this model is implemented on UML
domain which is described in next chapter.
3.1 User Profiles
UML-ALF consists of three main profiles. User profile, Tutor profile and Admin
profile. System admin can create other profiles. A new supportive agent is also created and
its knowledge level, settings are determined by this profile. Evaluation of users and whole
system can be monitored through the admin profile. Tutors can create course contents, exams
and practices for learners. User profile is the main profile of the framework. Necessary
information about user's knowledge level and preferences is gathered with the help of the
pre-tests and questionnaires when user signs up to the system. User may follow the learning
path which is assigned by the system or practice his/her knowledge in practice area.
3.2 User Modeling
While designing the User Model, modeling standards of IMS Learner Information
17
Packaging (IMS, 2001) and The Max Model (Lynch, Palmiter & Tilt, 1999) is based on.
3.2.1 User Model's User Profile:
User Profile is composed of two data structures as User Info and User Knowledge.
User Info has two main components from IMS Learner Information Packaging (IMS LIP) -
identification and security key -Identification stands for the learner information which
contains attributes such as user codes, names, addresses, contact information, status,
enrollment date, etc. Security key is used to keep the passwords and authentication keys of
user.
User Knowledge data structure has four main components from IMS LIP which are
relevant to this system. These components are Goal, Accessibility and Competency. Goal is
the description of learner's personal objectives. It is defined as learning path of user in this
system. Accessibility defines the cognitive, technical and physical preferences of the learner.
Lastly, competency is the skills that learner gained from the learning system.
3.2.2 User Model's Cognition Profile:
Cognition profile includes three main structures: Feedback Mechanism, Adaptive
Skills and Intelligence Skills.
Feedback mechanism which is explained in details in following sections is
connected with User Model by saving the records of feedbacks that have been transmitted to
user up to that time. Due to those feedback results and log analysis, user feedback
preference is gathered.
The statistics of adaptive methods – adaptive presentation, adaptive navigation
support and adaptive collaboration support – and intelligence methods – curriculum
sequencing, intelligent analysis of student solution, interactive problem solving support and
example-based problem solving support - that have been applied to user is also held in
cognition profile of user modeling.
Pretest score, posttest score, tutorial score, learning gain and other results of learner
are kept in evaluation structure of user modeling. Finally the result of questionnaires that
have been applied to users in order to get some feedback for our thesis is also a part of user
modeling.
3.3 Agent Modeling
Agent model is the subset of User Model. This model only applied to Supportive
18
agents among the agents in framework. Since supportive agents are counted as learners who
do not have any adaptive and intelligence skills, most of the design is inherited from User
Model except Cognitive profile. Agent Profile is composed of similar two data structures:
Agent Profile and Agent Knowledge. All the necessary data in Agent Profile is filled out
automatically when system admin creates a new Supportive agent. Initial knowledge level is
Figure 3.1 User Modeling
also defined during the creation process. There are two kinds of supportive agents according
to their goals. Their goal might be helping to the learner or hindering. There is no learning
path for supportive agents. They can learn by supporting or learn by collaborating. Learn by
supporting is the case when they give some hint to the learner. Learn by collaborating is the
19
situation that they learn from user's decision without helping them. In either way supportive
agent should join to the practice in order to improve its knowledge.
Figure 3.2 Agent Modeling
3.4 Framework
The framework can be abstracted as interface layer, coordination layer and data
layer. Interface layer is constituted with User profile, Feedback Mechanism, Supportive
agents and User Interface. Feedback mechanism generates effective feedbacks evaluating
user's actions in practice mode. Supportive agents help learner to achieve the exercises
successfully in practice mode. User Interface is the content viewer in education mode.
Content viewer is the place where learning materials are presented to the learner. It becomes
the exercise viewer in practice mode. Feedback mechanism and Supportive agents are fed by
20
Figure 3.3 Framework Layer
Diagnostic agent which is in coordination layer. User inputs in exercise viewer are
transferred through the Language Agent to the Resource. Language agent also provides the
communication between Resource in data layer and Diagnostic Agent. Adaptive agent is
responsible for displaying learning material according to the adaptive skills of learner. It gets
the information of learner with the help of Model Agent which has an access to User Model
in data layer. Adaptive agent also gets course data from data layer. Model agent's other duty
is to acquire supportive agent data from Agent Model. It sends this information to Diagnostic
Agent which maintains the learning of supportive agent.
3.5 AICC
Contents are presented in this study using AICC standards. The Aviation Industry
21
Computer-Based Training Committee (AICC) is an international association of technology-
based training professionals. AICC specifications are intended to be general purpose. Thus,
learning content manufacturers may produce their product for different branches of industry
at a lower cost. This is the main reason of AICC being so popular today regardless of being
in aviation or non-aviation industry. It is first created in 1988 by Aircraft companies in order
to standardize the learning materials. In 1993, AICC produced Computer Managed
Instruction which is widely accepted as the first runtime interoperability specification. It was
first used for local environment. With the arrival of the web, and widely spreading use of
browsers, in 1998 AICC added a browser-based format for transferring information in the
files. For web compatibility, AICC includes two main communication protocols. HTTP
AICC Communication Protocol (HACP) and Application Programming Interface (API). In
this research, HACP is chosen. In HACP the files of learning material are packaged as a web
page. Then, this package is posted to the server. The learning material can get information
from the server or send results to the server. Server responds the request of learning material
with plaint/text message as an AICC file. AICC files have a data structure in which
information of learning material is saved and stored. This data structure includes Core
(session id, score, date, exist status, lesson location), Core Lesson, Student Preferences,
Objective, Interactions and Paths.
3.6 Adaptive and Intelligent Techniques
In learning material, Adaptive Presentation technique is used in order to personalize
the content. According to Brusilovsky, P. (2001), adaptive presentation is grouped in three
categories: the adaptation of multimedia, the adaptation of modality and adaptive text
presentation. The adaptation of multimedia deals with manipulating the quality and size of
the multimedia content like video or image. The adaptation of modality is to select the media
to be used. This study focuses on the adaptive text presentation category of adaptive
presentation. The categorization of Brusilovsky, P. continues with two types of adaptive text
presentation: Natural language adaptation and canned text adaptation. Natural language
adaptation which is based on natural language understanding is beyond of this study. The
forms of canned text adaptation are used in UML-ALF. According to the characteristics of
users, the presentation of content is changed by adding or removing pages, hiding some parts
in different ways in this work. Due to the result of pre-test and values of user knowledge
structure, system creates a learning path which includes learning materials specifically
22
Figure 3.4 Workflow of the framework
selected for that user using curriculum sequencing skill. According to the knowledge base of
user, unnecessary topics are hidden from the learner on his learning path. With the help of
adaptive navigation support, system guides the user to follow his learning path easily using
direct guidance, link sorting and link annotation techniques.
While the user is being examined of his/her knowledge or making practice, system
responses a detailed feedback which is obtained from feedback mechanism to student using
intelligent analysis of student solutions skill. Also interactive problem solving support helps
user during the solution period of the problem following the same protocol with the help of
Supportive Agents. Supportive agents are agents which act like student teammates. They not
23
only respond to user's answers but also intervene during the solution process. Another
support while solving a problem is misleading hint support which are common mistakes of
other learner’s. Learners are informed that some agents might give misleading hints. Also,
tutor agent provides the correct hint using scaffolding approach after supportive agent’s
suggestion.
3.7 Intelligent Agents
Six different intelligent agents are used in this framework: Diagnostic Agent (DA),
Language Agent (LA), Feedback Mechanism (FM), Supportive Agents (SAs), Model Agent
(MA) and Adaptive Agent (AA). All of these agents are rational agents which does the right
thing. Here, doing the right thing is taking the right action which makes the agents most
successful. In order to measure the success, an evaluation is needed. Thus, the rationality of
an agent depends on four things (Russell & Norvig, 2003):
‐ Evaluation of success
‐ Initial knowledge of an agent
‐ Actions that an agent perform
‐ Perception of an agent
The task environment of an intelligent agent system is determined by combination of four
terms – Performance, Environment, Actuators and Sensors which is called PEAS shortly
(Russell & Norvig, 2003).
The task environment of this study is partially-observable (inaccessible). Since
each agent’s sensors have no access to complete state of environment. However, whole
agent system is fully-observable (accessible). It is stochastic (non-deterministic), because
one can never predict the behavior of users and tutors. Another property of the environment
is that it is sequential (non-episodic). Each step affects all future steps. Moreover it is a
dynamic environment. As the time flows, things change. On the other hand, time does not
flow continuously. It proceeds step by step. So, it is a discrete environment. The final
property of the environment is that it is a cooperative multi-agent environment. Agents are
not competitive, even they help each other.
3.7.1 Diagnostic Agent (DA):
Diagnostic agent is a model based reflex agent. It is the center of the intelligent
24
system. As the user makes progress through the interface, it gets translated rule package
from Language Agent (LA). Then it diagnoses the rule package by communicating resource.
It also can determine that user’s answer is an alternative solution to the question even it is
not stated in the resource. It can automatically corrects typos that user makes. Then it
generates reactions. These reactions may be a general hint, a crucial hint, a complete answer
or a misleading hint. Then sends those set of reactions to supportive agents (SAs). After user
decides to finish the practice, solution is translated to rule packages again by Language
Agent (LA). Then DA evaluates the rule packages and sends the results to the Feedback
Mechanism (FM).
3.7.2 Language Agent (LA):
Language agent is a simple reflex agent. The raw data from user interface of User or
Tutor profiles come to LA. It takes those raw data and translates them into meaningful –for
agents- rule packages. When user makes a progress or finishes the practice and tutor creates
a practice, the data is needed to be translated by LA.
3.7.3 Feedback Mechanism (FM):
Feedback mechanism is the structure where feedback messages are generated.
Feedback can be classified as positive and negative feedbacks. Negative feedbacks are
delivered in response to mistakes to help correcting them and positive feedbacks are
delivered in response to correct input. Those feedbacks might be verbal or non-verbal
feedbacks. Verbal feedback can be written or spoken whereas non-verbal feedbacks are
generally mimics, gestures, sounds or pictures.
There are 3 kinds of feedbacks: Failure feedback, Success feedback and warning
feedback. Failure feedback is not a simple feedback message. For a detailed feedback
message that includes some hint to solve the problem, neural networks are used. Especially
success feedback messages consist of both verbal and non-verbal feedbacks.
Warning feedbacks are neither failure nor success messages. In fact, they are not
related with the result of solution. Warning feedbacks are generated when system could not
understand or fulfill the request of the user. Also they might be system error,
incompatibilities, improper usage, etc.
Feedback Mechanism is also a simple reflex agent. After user finishes the practice,
consecutively LA translates the raw data to rule packages. Then DA diagnoses the rules and
sends results to FM. If the result is correct, it returns a positive feedback. Otherwise, it
25
Table 3.1 Description of the task environment
Agent Type Performance
Measure
Environment Actuators Sensors
Diagnostic
Agent (DA)
successful
reactions,
misleading
reactions, auto
completion,
alternative
solutions
Resource, FM,
LA, SAs, User
check data,
generate
reaction, send
feedback data
rule package (LA),
finish practice (User)
Language
Agent (LA)
translation User, Tutor,
DA, Resource
translate raw input data (User,
Tutor)
Feedback
Mechanism
(FM)
positive
feedback,
negative
feedback
DA, User response
feedback
feedback data (DA)
Supportive
Agents
(SAs)
misleading
response, no
response,
helpful
response
User, DA give hint ask for help (User),
generate reaction
(DA)
Model
Agent (MA)
initializing
user model,
initializing
agent model
User, Admin,
User Model,
Agent Model,
DA, AA
create user
profile, create
SA, response
User Model,
response
Agent Model
sign up to the system
(User), create SA
(Admin), get User
Model (AA), get
Agent Model (DA)
Adaptive
Agent (AA)
adapting
learning
material
User, Course,
MA
send course
content, get
course raw
data, get user
model
get course content
(User), send course
raw data(Course),
send user model
(AA)
26
examines the data and creates an intelligent negative feedback, which tells user clearly where
he/she made a mistake. Feedback mechanism has three outputs: sound, gesture and text.
Except the text it produces, it also makes a sound effect and responses a proper gesture.
3.7.4 Supportive Agents (SAs)
Supportive agents are agents which act like student teammates. They not only
respond to user's answers but also intervene during the solution process. Basically they have
three main actions. They can help by giving useful suggestions. Those suggestions may be
crucial or might not help at all. Secondly, they may not interfere or they have no opinion
about that situation. Lastly, they can hinder and convince user with their opinions. All of
these are possible teammate behaviors in collaborative work. User may force the irrelevant
SA to comment on subject. If it has knowledge about subject, may give correct information
or if it has no idea, it make up an answer which is not correct. User also may force other SAs
- helping SA or hindering SA- to stop or to keep on giving advices. The actions of simulated
collaborative students – SAs - are modeled using artificial intelligence methods such as
Neural Networks and Hidden Markov Model.
Learning from mistakes is one of the learning methodologies. Trying to mislead
learner with a common mistake is also used in (Mengelle et al., 1998). The benefit of
learning by making mistakes is also discussed in (Dweck, 2007; Izzo & Mazzone, 2008;
Chialvo & Bak, 1999). As a similar approach, “trial and error”, or “trial by error” or “try an
error”, is a general method of problem solving, fixing things. In the field of computer
science, the method is called “generate and test”. In elementary algebra, when solving
equations, it is "guess and check". Therefore, it is decided to implement suggesting a
common mistake by supportive agents as an alternative option.
SAs are goal-based agents. They may have two goals: misleading the user or helping
user. When user asks for help, related SA gets the reactions from DA and decides its action
among the reactions. It automatically gives the hint to the user. However, when user makes
a progress, each SA may have an opinion about that move, but cannot tell directly what it
thinks. Instead, it blinks which means it has something to tell. User may allow telling its
opinion or not.
3.7.5 Model Agent (MA)
Model agent (MA) is a model based reflex agent. When user signs in to the system at
first time, user encounters pre-tests and questionnaires. Due to the results of these tests,
Model Agent initializes the model of the user. It sets up the knowledge level of the user,
27
determines his/her preferences and creates his/her learning path. When Adaptive Agent
displays the learning material to the user, it asks MA for the necessary information about the
user model and MA sends the information back. Another duty of MA is to initialize the agent
model for the supportive agent after system admin determines the information of newly
created SA. While supporting user in practice area, Diagnostic Agent asks MA for related
SA's agent model and MA sends the information back.
3.7.6 Adaptive Agent (AA)
Adaptive Agent (AA) is a model based reflex agent. The main duty of AA is to
display the learning material after adapting it according to the user's user model. It gets the
necessary information of user from MA.
28
CHAPTER 4
METHOD
4.1 Implementation & A Case Study: UML
4.1.1 Unified Modeling Language (UML) Domain
As software projects become larger, programming complexity will increase.
Moreover, the number of programmers in the project will increase as well. Therefore, a
necessity for standard modeling and design language appears. Unified Modeling Language
(UML) is a collection of methods which explain and define how to model work flows, rather
than being a programming or software development language. It is standardized and created
by Object Management Group.
The UML model and the set of diagrams of a system should be differentiated. A
diagram is one of the parts of system model which is visually represented. There are totally
14 diagrams in UML 2.2. Those diagrams are grouped under two main categories: Structure
(Static) Diagram and Behavior (Dynamic) Diagram. Structure Diagrams and their features
are shortly described as:
Class diagram: is one of the most used UML diagram. It aims to represent classes that
constitute the basis of Object Oriented Analysis, Design and Programming.
Component diagram: It shows the components that constitutes the software system and
dependencies among other components.
Composite structure diagram: describes the internal structure of a class and the
collaborations that this structure makes possible.
Deployment diagram: serves to model the hardware used in system implementations, and
the execution environments and artifacts deployed on the hardware.
Object diagram: shows a complete or partial view of the structure of a modeled system at a
specific time.
29
Package diagram: depicts how a system is split up into logical groupings by showing the
dependencies among these groupings.
Profile diagram: operates at the meta model level to show stereotypes as classes with the
<<stereotype>> stereotype, and profiles as packages with the <<profile>> stereotype. The
extension relation (solid line with closed, filled arrowhead) indicates what meta model
element a given stereotype is extending.
Behavior diagrams have following diagrams:
Activity diagram: represents the business and operational step-by-step workflows of
components in a system. An activity diagram shows the overall flow of control.
State machine diagram: standardized notation to describe many systems, from computer
programs to business processes.
Use case diagram: shows the functionality provided by a system in terms of actors, their
goals represented as use cases, and any dependencies among those use cases.
Since behavior diagrams illustrate the behavior of a system, they are used extensively to
describe the functionality of software systems.
Interaction diagrams: a subset of behavior diagrams, emphasize the flow of control and
data among the things in the system being modeled:
• Communication diagram: shows the interactions between objects or parts in terms of
sequenced messages. They represent a combination of information taken from Class,
Sequence, and Use Case Diagrams describing both the static structure and dynamic
behavior of a system.
• Interaction overview diagram: is a type of activity diagram in which the nodes
represent interaction diagrams.
• Sequence diagram: shows how objects communicate with each other in terms of a
sequence of messages. Also indicates the lifespan of objects relative to those messages.
• Timing diagrams: Is a specific type of interaction diagram, where the focus is on timing
constraints.
4.1.2 Training Process of UML-ALF
The training process of UML-ALF starts with registration of user to the system. User
information, additional preferences and settings for experimental procedure is defined during
the registration.
System is proposed of two main parts.
30
Book image represents the Study Part of the system. This part is managed and
applied with the help of the adaptive techniques which are introduced in previous
chapter. User can take online courses and solve post-tests which are adapted
according to their characteristics and knowledge level.
Pencil image represents the Practice Part of the system. This part is managed and
applied with the help of the intelligent techniques. Three UML diagrams – Class
Diagram, Use – Case Diagram and Activity Diagram- are chosen as practice
models. Users can improve their own skills with the help of this interactive practice
screen, which can suggest new ideas, evaluate and track user moves in very close
manner.
Figure 4.1 UML-ALF Home Page
After signing to the system, at first time user is introduced with pre-test (Figure 4.4). According to the result of pre-test, a proper learning path (Figure 4.3) is assigned to the user
31
Figure 4.2 UML-ALF Registration page
by the adaptive agent. A training package includes e-learning material, post-test and
intelligent practice environment. User knowledge model determines the courses that user
may attend. Green color shows the training packages which was already completed. Orange
color shows the unlocked materials whereas Red color implies the locked materials. (Figure
4.5)
Figure 4.3 UML-ALF Learning path
32
Figure 4.4 UML-ALF pre-test
UML-ALF has eight chapters while training the UML domain: Introduction, Object
Oriented Modeling, Class and Object Diagrams, Use Case Diagrams, Component and
Deployment Diagrams, Sequence and Collaboration Diagrams, State Diagrams and Activity
Diagrams. Each chapter is formed of an e-learning course which is designed by following
AICC standards and a post-test which evaluates the knowledge gain of user from online
course. Also three chapters - Class Diagram, Use – Case Diagram and Activity Diagram –
has an additional practice link that redirects to Practice Part of the system. E-learning
materials are also prepared with adaptive techniques. If user has mastered some parts of the
chapter but not the whole, then when he/she experiences the same course second time, only
the parts which were not fully learned before are introduced to the user.
There is no limitation for taking courses over and over again. If the user has already
completed the course before, system warns the user about whether taking the course with full
parts (without adaption) or not. Detailed information about the process of taking e-learning
materials is logged and related charts can be produced in admin profile of the system.
Practice part of the system uses intelligent techniques of artificial intelligence. Users
may improve their knowledge level by practicing the subjects learned in Study Part. In
33
practice screen user is asked to generate a UML diagram using the information given on top
of screen (Figure 4.7 item-3). Learners often demand feedback, tutorial and guidance in
learning environments. Generally these requirements are being provided with the help of
human–computer interaction in computer-based learning environments. (Kızılkaya & Aşkar,
2008). There are three supportive agents that help user during the practice period (Figure 4.7
item-10). Two of these agents always give the correct suggestions as long as they know the
Figure 4.5 UML-ALF Course Tree
subject. However one of them suggests a common mistake which has been made in that
practice before. Last agent is a tutor who stands at the right-top of the screen (Figure 4.7
item-9). When an agent makes a comment about a move that user made or suggests a new
idea, tutor evaluates this information and gives the most correct and definite result to the
learner. Firstly, user decides which UML diagram is used for that practice, then from toolbar
(Figure 4.7 item-7) at the left side of the screen, he/she chooses the necessary elements.
Those elements are placed to the practice area by drag & dropping. If it is an element that
34
Figure 4.6 First page of Object Oriented Modeling Course
can be named, user should change the default name of the element to the correct name
according to the scenario of the practice. With the relation tools, like directional,
aggregation, composition, etc. the relation between the elements can be provided. Also, if
necessary, cardinality of the relation may be defined.
After each move of the learner, supportive agents may comment about that move.
They indicate their desire of making comment by blinking with gray colors. After they show
their will, user may click on agent’s photo to learn about their idea. They express their
opinion with a speech balloon. Finally, tutor agent makes a decision about whether the
expression of supportive agent is true or not. He shows his decision by nodding his head and
with the color of his picture. If he nods his head left to right and the color is red then he
means the expression is not correct. If he nods his head up and down and he color is green
then he means the expression is correct.
35
Figure 4.7 UML-ALF Practice Screen
Dragging an element clears its relations with the other elements and dragging an
element to the recycle bin disposes it. The icon that is numbered with 5 in figure 4.7 clears
whole practice area. After learner decides that he/she is done with the practice, he/she
presses the ‘Finish Practice’ button to learn the result of the practice. The score of user and
the correct solution of the practice is shown on a popup window. Supportive agents have
their own knowledge model. Therefore, as they made new comments and attend new
practices, they improve their knowledge level. There are totally 2 tutor agent and 10
supportive agents. Since they are goal-based agents, they are classified due to their goals.
Three of them aim to frustrate the learner and rest of them are helpful agents.
Every move of learner is being tracked and logged by the system in order to evaluate
precisely. When learner makes same mistakes again and again, system thinks that there
36
would be a problem and asks a relevant question to the learner. If learner answers this
question correctly, he/she goes on practice; otherwise a summary of the relevant topic is
shown to the learner. Then, another question is asked. If learner still cannot find the correct
answer, finally he/she is redirected to the course of the subject.
4.1.3 Building Resources
The resource is in charge of the diagnosis of student knowledge and informs the
actors about what they can do next. So, clearly, the global quality of the ITS also depends on
the expertise of the resources. Thus, it is decided to build a resource structure.
Figure 4.8 UML-ALF User Tracing Flow Chart
37
The resource structure consists of three main parts:
• Subject Information: an identification of resource, UML diagram type, the list of
related subjects and question text.
• Answer Rules: it is the whole solution of the practice. It is initially determined by
the designer of the practice.
• Practice Hints: a set of hints regarding related subjects.
Tutor is the designer of a practice. The resource can be inserted to the system either
through user interface or via an XML file. An example of a resource structure data can be
found in Table-4.1 and its output is stated in Figure 4.9.
Table 4.1 Resource structure
Resource ID 1 UML Diagram Type Class Diagram Related Subjects <class_diagram>,<directional>,<aggregation>,<cardinality>
Question Text
A company consists of departments and employees. A department has responsibility for zero, one or more projects. A company has: name, address and telephone. A department has name. A project has name. An employee has name, address, a social security number (SSN) and is participating in zero, one or more projects.
Answer Rules
[Employee|name;address;SSN] [Company|name;address;phone] [Project|name] [Department|name] [Company] + 1 - > * [Employee] [Company] + 1 - > * [Department] [Department] – 0... * > [Project] [Project] – 0... * > [Employee]
Hints Actors are not an element of Class Diagrams. The relationship between those subjects is not ‘Composition’. Watch out the cardinalities!
38
4.2 Definition of Propositions
1. Supportive agents provide contribution to the learning achievement of learners in
distance education.
2. Learners are interacted with supportive agents as it is purposed.
2.1 The methodology that agents give misleading hints which are common
mistakes of other learners had positive effects on learning the subject.
2.2 The methodology which forces learners to renew their knowledge as they
make mistakes is effective on their learning achievement.
3. Intelligent agents and computer based tutors can give effective feedback to the
learners which is close to the performance of human tutors.
Figure 4.9 Output of the resource structure stated in Table 4.1 (Harris, 2009)
4.3 Participants of the study
In order to assess that UML-ALF has accomplished its objectives, it is decided to
follow an experimental procedure. Three different features of this framework were
considered during the experiment. Those features are adaptive hypermedia, intelligent
tutoring system and advanced feedback mechanism.
Currently, the proposed UML-ALF is available on the web to provide personalized
UML learning. In order to verify the quality of UML-ALF, people were invited to test this
system. Purposive sampling method (Patton, 2002) was preferred to choose the suitable
sample for our experiment. The main idea is to gather two groups one of which is experts of
the subject and the other one is willing to learn the subject but has no background.
39
Before testing the system, participants joined a survey [Appendix B] that determines
in which category they should be included. According to the results of survey people are
gathered in two groups. These groups are named as Experienced and Inexperienced. People
in experienced group has advanced knowledge about UML, they took a UML related course
in their education, use the benefits of UML diagrams in their current work. On the contrary,
participants in inexperienced group are lack of UML knowledge or little if any, however they
are interested in subject. In each category, participants were divided into two sample groups.
While first group had no benefit of UML-ALF's features, second group has taken advantage
of those features. A total of 32 participants were analyzed and reported within the scope of
this study. Demographics of the participants were not taken into consideration, since the
experiment is only interested in learning background of the participants.
4.4 Data Collection Procedure
When carrying out a study where data acquisition is needed from the human
participants, an approval from the Applied Ethics Research Center is obligatory at METU.
Therefore, within the scope of this study, an ethical consent is prepared and an approval from
the ethics committee is obtained before data gathering process. (Appendix E). Each
participant in this research has signed the ethical consent before participation.
Before performing the test to whole sample, a pilot study is conducted. The
pilot study is carried out with 19 individuals. Participants joined the survey [Appendix B]
that determines in which category they should be included. According to this survey;
Experienced Group has 7, inexperienced group has 12 participants. Those participants are
gathered from the calls that are announced in newsgroups of Computer Engineering
Departments of the METU and IYTE and other communication newsgroups and forums.
Participants made use of all system features. All courses and one practice were available
during pilot test. The aim of the pilot study was to simulate the data gathering process and to
foresee any problems that might be faced during data gathering.
The results of the pilot test was evaluated in two parts. Table 5.1 shows the results of
adaptive tests. The number of participants answered the test, the number of minimum and
maximum correct answers, median and standard deviation for each chapter are stated in
Table 5.1. According to the this table number of participants were decreased as time goes by.
It is concluded that participants has lost their interest. It is decided that one major cause of
this might be the high threshold that is defined for moving from one unit to the next. The
40
Table 4.2 Adaptive Results of Pilot Test
Test Number of
Participants
min max M SD
Introduction pre-test 19 1 8 4,16 2,03
Introduction post-test 17 2 10 6,82 2,04
Object Oriented Modeling pre-
test 10 1 8 5,6 2,33
Object Oriented Modeling post-
test 9 3 9 7,56 1,77
Class and Object Diagrams post-
test 8 3 7 5,38 1,32
Class and Object Diagrams pre-
test 8 4 9 7,13 1,62
Use Case Diagrams pre-test 6 2 8 5,65 1,87
Use Case Diagrams post-test 4 4 9 6,9 1,88
Component and Deployment
Diagram pre-test 5 4 8 6 1,41
Component and Deployment
Diagram pre-test 4 4 10 7,5 2,18
Sequence and Collaboration
Diagrams pre-test 2 3 4 3,5 0,5
Sequence and Collaboration
Diagrams pre-test 2 5 10 7,5 2,5
State Diagrams pre-test 3 2 6 4 1,63
State Diagrams post-test 1 6 8 7,34 0,94
Activity Diagrams pre-test 1 4 4 4 0
Activity Diagrams post-test 1 6 6 6 0
score threshold was lowered from 85% success to approximately 70% success for each unit.
The results of practice part of the system showed that related parameters for
41
evaluating the success of intelligent agent techniques such as logging learner’s moves and
agent’s responses did not constitute enough data for the researcher. Thus, it is decided to
observe participant’s moves in Human Computer Interaction laboratory and one-to-one
experiments.
4.5 Data Analysis Methodology
As previously stated, most e-learning systems focus on intelligent tutoring
methodologies, but ignore basic learning theories. However, including these theories to
learning environments is highly crucial, since they provide clearer images of the students’
learning circumstances. Therefore, learning systems could adapt according to learner’s
characteristics. Keller & Sherman (1974) proposed a study model called Personalized
System of Instruction (PSI) for school education which allows students to follow a lecture at
an individual speed. System divides the whole content into basic learning elements and each
of these units has a content and a test. A score threshold is defined for proceeding to the next
unit and a student must get enough score to move from one unit to the next. In our study,
Keller’s Personalized System was adopted.
In most evaluation systems, student’s score is set according to the major assessment
rules. Another factor regarding quality of learning is learning achievement. It is stated that
Table 4.3 Content of the Implemented Course UML
UML
Chap. Subject Chap. Subject
1 Introduction 5 Component and Deployment
Diagram
2 Object Oriented Modeling 6 Sequence and Collaboration
Diagrams
3 Class and Object Diagrams 7 State Diagrams
4 Use Case Diagrams 8 Activity Diagrams
42
Table 4.4 Snapshot of the Learning Nodes of Chapter 3
Chapter 3 – Class and Object Diagrams
Topic ID Contents
1 Classes and Objects
2 Associations and Links
…
7 Packages and Subsystems
most effective learning performance is determined by using assessments and test (Scott,
Beichner, Titus & Martin, 2000; Shih, Lin & Change, 2003). The assessment or test
generally produce two consequences. The first one lets the learning system or tutors analyze
the progress in achieving the learning objective, and the second one gives effective feedback
to the learners. However, the conventional test theory does not respond easily to individual
performance. In this study, the modern test theory is applied in order to obtain each student’s
learning achievement.
The practice part of the UML-ALF has been evaluated in two ways. 5 participants
has been experienced UML-ALF in Human Computer Interaction Laboratory in METU. The
purpose of this experiment was to research the nature of interaction between participants and
UML-ALF in order to enhance the design and utilization of more usable and humanly
acceptable system. Also, the correlation between supportive agents and student’s learning
achievement was observed.
In practice part of UML-ALF, the relationship between two variables, the usage of
supportive agents by the learner and their practice score was observed. The hypothesis is that
how much learner made use of supportive agents affects their practice score. Here, there is
no need to worry about the direction of causality, since it’s not likely that practice score
causes supportive agent usage. Supportive agent usage is a number which increases by 1
each time the learner get helped.
43
CHAPTER 5
RESULTS
The purpose of this study was to present an agent-based adaptive learning
framework whose goal is to implement effective tutoring system with the help of AI
techniques. In this section, the results for each research questions are introduced.
5.1 Proposition 1
First proposition is defined as supportive agents provide contribution to the learning
achievement of learners in distance education. In order to find the contribution of supportive
agents, the correlation between two variables, the usage of supportive agents by the learner
and their practice score was based on.
The hypothesis is that how much learner made use of supportive agents affects their
practice score. Here, there is no need to worry about the direction of causality, since it’s not
likely that practice score causes supportive agent usage.
Table 5.1 Descriptive Statistics of Practice Results
Variable Mean Standard Deviation
Variance Sum Min. Value
Max. Value
Range
Score 87,32 10,26 105,27 2707 62 100 38
Ias 15,32 6,02 36,24 475 5 29 24
44
Table 5.2 Statistics for correlation between # of Agent Suggestion and Practice Score
Participant# of Agent
Suggestion (x)Practice Score (y) x*y x*x y*y
1 29 100 2900 841 10000 2 19 81 1539 361 6561 3 26 92 2392 676 8464 4 11 76 836 121 5776 5 22 92 2024 484 8464 6 24 100 2400 576 10000 7 19 96 1824 361 9216 8 18 94 1692 324 8836 9 6 90 540 36 8100 10 13 100 1300 169 10000 11 15 88 1320 225 7744 12 7 66 462 49 4356 13 17 80 1360 289 6400 14 6 62 372 36 3844 15 16 80 1280 256 6400 16 16 96 1536 256 9216 17 8 72 576 65 5184 18 12 92 1104 144 8464 19 18 96 1728 324 9216 20 12 82 984 144 6724 21 16 90 1440 256 8100 22 21 98 2058 441 9604 23 18 86 1548 324 7396 24 9 76 684 81 5776 25 5 72 360 25 5184 26 18 90 1620 324 8100 27 14 94 1316 196 8836 28 22 100 2200 484 10000 29 14 96 1344 196 9216 30 7 82 574 49 6724 31 17 88 1496 289 7744
Sum= 475 2707 42809 8402 239645
45
Participants that made use the benefits of the intelligent techniques of UML-ALF
has produced totally 31 practice results in 3 practices ( Class Diagrams, Use Case Diagrams
and Activity Diagrams). Symbol ‘r’ is used to stand for the correlation. According to the
result of Table 5.6, r is calculated as 0.69. The result of r shows that there is a positive
correlation between variables practice score and number of agent suggestion.
Once the correlation was computed, the probability that the observed correlation
occurred by chance was determined. That is, a significance test is conducted. In this case, the
mutually exclusive hypotheses: null hypothesis and alternative hypothesis was tested.
The significance level is used as common significance level of alpha = 0.05. This means that
a test was conducted where the probability that the correlation is a chance occurrence is no
more than 5 out of 100. The degrees of freedom (df) is simply equal to N-2, in this study it is
29. Finally, it is decided to do two-tailed test, since it is not clear that the relationship
between practice score and number of agent suggestions would be positive or negative. With
these three parameters - alpha = 0.05, df = 29 and type of test is two-tailed, using the lookup
table of critical values for Pearson’s r, the critical value is found as 0.355. This means that if
Figure 5.1 Scatter Diagram for Practice Score and Agent Suggestion
46
the correlation value is greater than 0.355 or less than -0.355, it can be concluded that the
quite a bit higher, it can be concluded that it is not a chance finding and that the correlation is
"statistically significant" (given the parameters of the test). The null hypothesis was rejected
and the alternative was accepted.
5.2 Proposition 2
Practice part of the UML-ALF also has been experimented in Human Computer
Interaction Laboratory. There is a single eye tracker (Tobii 1750) which collects data on
where on the screen the user looks, how long a glance is, how many times the user looks at a
certain point on the screen, and all the eye movements of the subject during the experiment is
also monitored. Tobii Studio is software developed by the manufacturers of the eye tracker
that transforms the information received from the reflectors, the infrared detector cameras
into visual and digital data, records them and later provides various tools for analyzing the
data saved. Five learners participated in the experiment. As three participants completed
three of the practices, two participants completed just one practice. The observations during
the experiments showed that supportive agents have a significant effect on completing the
Figure 5.2 Time to First Fixation
47
practice and learning the subject. The heat map results of HCI lab experiments including
mouse clicks are presented in Appendix D. In this map, the fixations are represented
according to their time and number as a degrading pattern (red color represents the areas
which are looked longer and most). By means of these output, at a first glance, the
information about the areas which the participants look most is obtained. It is indicated that
all participants were interested with supportive agents and they completed the practices
successfully. Most of the participants’ interaction data are approximately 80% valid, whereas
participant 2 had about 55%. It is assumed that the reason for the shortfall of data gathering
is that he was wearing glasses. And it is believed that because of this reason, there was not as
much intensity on supportive agents as other participants on his heat map results.
Another evaluation criterion is the statistics of area of interests (AOI). The practice
screen of UML-ALF is divided into four parts which are called as area of interest in Tobii
Studio. First area is the right side of the screen where four agents exist. Second area is the
left side of the screen that tools are presented. Third area is the board which is the main
screen the user makes practice. Fourth area is the top of the screen where the text of practice
exists.
Figure 5.3 Fixation Count
48
Five filters were used to collect the statistics of area of interests: time to first
fixation, fixation duration, fixation count, visit duration and visit count. Figure 5.2 shows the
first fixation times of each areas. It is clearly seen that mostly participants first look at board
area since it is the center of the screen. Then, learners focus on the practice text.
Following area is the tools area where they start their practice. Lastly, participants
glance the agents. It is believed that this is the ordinary path participants would follow.
When evaluating the results of statistics for agent area, blinking agents and their responses
with short sentences should be considered as the duration of their fixation and visit
parameters would not last long. Figure 5.3 presents the fixation parameters of the AOI.
According to these figures, board is the area where participant most focused on. It is not
expected that the agent area would be the most focused part among others, since it is a place
where participants look for short periods. However, the result of fixation times shows that
the agent part is used nearly as much as text and tools areas.
According to 5.4 which introduces the visit parameters of the AOI, the board area is
again the most visited part. The result of these figures also confirms the results of fixation
parameters. The visit duration and count values for the agent area is also as much as text and
tools areas.
Mouse clicks are accepted as verification that participants get suggestion from
supportive agents. Left mouse clicks on agent areas are illustrated in screenshots with mouse
icons. (Appendix E)
Total fixation count, visit count and number of mouse clicks are shown in Figure
5.5. Average values of fixation counts, visit counts and number of mouse clicks are 89.09, 51
and 39.09 respectively. Fixation count implies number of times the participant fixates on
AOI. Visit count is the number of visits within an AOI. The difference between fixation
count and visit count is that when participant fixates on some point then in another point in
same AOI, it is counted as 2 for fixation count and 1 for visit count. So, if visit counts and
mouse clicks are equal, then it means that each time participant looked at agent area, he/she
get suggestion from Supportive Agents. However, when participant looks at blinking agents
and he/she might not click. According to the average results of HCI experiments (Figure
5.5), it is concluded that with a 77% probability, participants get help from Supportive
Agents when they look at them.
As a consequence, it is accepted that participants are interacted with supportive
agents due to the results of HCI experiments. Second proposition has already showed that the
correlation of the participant’s score with supportive agent’s suggestion.
49
In practice part of the UML-ALF, some agents may give misleading hints which are
common mistakes of other learners. In order to decide whether this methodology contributes
to learning achievement of the user, a question was asked in evaluation survey. According to
the result of the qualitative research (Figure 5.26) 56% of the participants agree with the idea
of supportive agents that give misleading hints has positive effect on their learning process,
whereas 19% of the participants reject the idea.
Another methodology used in practice part was tracking learner’s movements. When
learner makes same mistakes again and again, system thinks that there would be a problem
Figure 5.4 Visit Count
and asks a relevant question to the learner. If learner answers this question correctly, he/she
goes on practice; otherwise a summary of the relevant topic is shown to the learner. Then,
another question is asked. If learner still cannot find the correct answer, finally he/she is
50
Figure 5.5 Total Fixation, Visit Count and Mouse Clicks for Agent Area
Figure 5.6 Scatter Diagram between Practice Score and Number of Mistakes
51
Table 5.3 Statistics for correlation between # of mistakes and Practice Score
Participants# of
mistakes (x) score (y) x*y x*x y*y 1 10 100 1000 100 10000 2 1 81 81 1 6561 3 4 92 368 16 8464 4 13 76 988 169 5776 5 7 92 644 49 8464 6 2 100 200 4 10000 7 16 96 1536 256 9216 8 2 94 188 4 8836 9 0 90 0 0 8100 10 3 100 300 9 10000 11 1 88 88 1 7744 12 7 66 462 49 4356 13 5 80 400 25 6400 14 6 62 372 36 3844 15 2 80 160 4 6400 16 1 96 96 1 9216 17 8 72 576 64 5184 18 1 92 92 1 8464 19 4 96 384 16 9216 20 6 82 492 36 6724 21 3 90 270 9 8100 22 2 98 196 4 9604 23 5 86 430 25 7396 24 6 76 456 36 5776 25 5 72 360 25 5184 26 9 90 810 81 8100 27 1 94 94 1 8836 28 22 100 2200 484 10000 29 5 96 480 25 9216 30 7 82 574 49 6724 31 11 88 968 121 7744
Sum= 175 2707 15265 1701 7327849
52
redirected to the course of the subject. The correlation between two variables, the usage of
supportive agents by the learner and their practice score was based on. In order to evaluate
the methodology that tracks user movements, a correlation between number of mistakes and
practice score is examined. The hypothesis is that how much learner made a mistake affects
their practice score. Participants that made use the benefits of the intelligent techniques of
UML-ALF has produced totally 31 practice results in 3 practices ( Class Diagrams, Use Case
Diagrams and Activity Diagrams). According to the result of Table 5.7, r is calculated as -
0.0002. The result of r and Figure 5.6 show that there is no correlation between variables
practice score and number of mistakes. However, according to the Table 5.7, it is also seen
that participants that made mistakes also get high scores from practices. It can be concluded
that the methodology of tracking users has the ability of teaching subject to the
inexperienced users and increasing their learning achievement.
Also, a question was asked about tracking user movements in evaluation survey. According
to the qualitative research (Figure 5.24) 47% of the participants agree with the idea of
tracking user movements and redirecting them to content materials, whereas 44% of the
participants reject the idea. The reason for the high percentage of disagreement might not be
related with the failure of the methodology, on the contrary it can be related to exhausting
effects of redirecting user from practice screen.
5.3 Proposition 3
Third proposition is introduced as intelligent agents and computer based tutors can
give effective feedback to the learners which is close to the performance of human tutors.
The result of this proposition is determined by the evaluation surveys which were filled out
by participants after experiencing UML-ALF. According to the results of these surveys, the
feedback which is given by intelligent agent and computer based tutors in this study is
satisfactory; however it is still not so close to human tutors. (Figure 5.20)
Since it is a relative parameter that feedback is effective or not, in order to make
more accurate and quantitative decisions, in future works, different learning platforms can be
compared with a larger sample of participants.
5.4 Survey Results
Results of each question in evaluation survey [Appendix C] are presented in pie
charts. The results of evaluation survey show that participants find the content material
53
sufficient in terms of information and visuality. It is also verified that adaptive techniques
that were implemented in study part of UML-ALF was helpful. Most of the participants find
the learning path convenient which is assigned by the system. It is stated that supportive
agents' help was valuable. They agree that practice screens help to test and improve their
knowledge about the subject. However, according to the participants, the feedback generated
in practice part was not enough. As a consequence, they think that system is successful in
teaching UML.
Figure 5.7 Result of Question 1 in Evaluation Survey
Figure 5.8 Result of Question 2 in Evaluation Survey
54
Figure 5.9 Result of Question 3 in Evaluation Survey
Figure 5.10 Result of Question 4 in Evaluation Survey
Figure 5.11 Result of Question 5 in Evaluation Survey
55
Figure 5.12 Result of Question 6 in Evaluation Survey
Figure 5.13 Result of Question 7 in Evaluation Survey
Figure 5.14 Result of Question 8 in Evaluation Survey
56
Figure 5.15 Result of Question 9 in Evaluation Survey
Figure 5.16 Result of Question 10 in Evaluation Survey
Figure 5.17 Result of Question 11 in Evaluation Survey
57
Figure 5.18 Result of Question 12 in Evaluation Survey
Figure 5.19 Result of Question 13 in Evaluation Survey
Figure 5.20 Result of Question 14 in Evaluation Survey
58
Figure 5.21 Result of Question 15 in Evaluation Survey
Figure 5.22 Result of Question 16 in Evaluation Survey
Figure 5.23 Result of Question 17 in Evaluation Survey
59
Figure 5.24 Result of Question 18 in Evaluation Survey
Figure 5.25 Result of Question 19 in Evaluation Survey
Figure 5.26 Result of Question 20 in Evaluation Survey
60
Figure 5.27 Result of Question 21 in Evaluation Survey
Figure 5.28 Result of Question 22 in Evaluation Survey
61
CHAPTER 6
DISCUSSION
In order to assess that UML-ALF has accomplished its objectives, we followed an
experimental procedure. Three different features of this framework were considered during
the experiment. Those features are contribution of supportive agents, interaction between
SAs and learners, and advanced feedback mechanism.
6.1 The Contribution of Supportive Agents
The correlation coefficient r = 0.69 shows that there is a positive correlation between
variables practice score and number of agent suggestion. This means, as the participants
benefit from supportive agents, they get higher scores. With the significant test, it is
concluded that the value is not a chance finding and that the correlation is "statistically
significant". The null hypothesis was rejected and the alternative was accepted.
The participants of the study were examined in terms of their professions. According
to the results of survey [Appendix B] participants are gathered in two groups. These groups
are named as Experienced and Inexperienced. People in experienced group has advanced
knowledge about UML, they took a UML related course in their education, use the benefits
of UML diagrams in their current work. On the contrary, participants in inexperienced group
are lack of UML knowledge or little if any, however they are interested in subject. The
correlation coefficient of inexperienced group was 0.81 whereas it is 0.62 for experienced
group. This result implies that there is more linear relationship between practice score and
number of agent suggestion for inexperienced group than experienced group. It can be
concluded that inexperience group had much to learn and benefit from supportive agents
more than experienced group.
A similar work which is an adaptive educational hypermedia system called
MATHEMA (Papadimitriou et al., 2009) also supports curriculum sequencing, adaptive
62
presentation, adaptive navigation and interactive problem solving techniques. It is stated that
an interactive tutor cannot be implemented in their work. The agents sleeps from one help
request to another. On the contrary, in UML-ALF supportive agents response to each user
move during their learning period. Since domains and evaluation methods are different for
both works, the comparison of the success of frameworks cannot be precise.
6.2 The Interaction between Supportive Agents and Learners
6.2.1 HCI experiments
Practice part of the UML-ALF also has been experimented in Human Computer
Interaction Laboratory. Mouse clicks are accepted as verification that participants get
suggestion from supportive agents. Left mouse clicks on agent areas are illustrated in
screenshots with mouse icons (Appendix E). Average values of fixation counts, visit
counts and number of mouse clicks are 89.09, 51 and 39.09 respectively. Fixation count
implies number of times the participant fixates on AOI. Visit count is the number of visits
within an AOI. The difference between fixation count and visit count is that when
participant fixates on some point then in another point in same AOI, it is counted as 2 for
fixation count and 1 for visit count. So, if visit counts and mouse clicks are equal, then it
means that each time participant looked at agent area, he/she get suggestion from Supportive
Agents. However, when participant looks at blinking agents and he/she might not click.
According to the average results of HCI experiments, it is concluded that with a 77%
probability, participants get help from Supportive Agents when they look at them. It is a high
usage rate for SAs. Therefore, correlation coefficient result shows that supportive agents
play important role for increasing learning achievement of the user. In addition to this, HCI
experiment results verify this conclusion with a high usage rate.
6.2.2 Suggesting Common Mistakes as Misleading Hints
The communication protocol of the agents and building the practice resources are
based on the study of Mengelle et al. (1998). However, since domains of these studies are
different, techniques for building the resource are also different. Language agent is
responsible for building and interpreting the resource in UML-ALF. Mengelle et al. (1998)
focus on modeling of pedagogical expertise and actors’ learning abilities whereas in our
study the purpose was to implement these methodologies. The idea of generating misleading
hints is originated from Mengelle's work. But this idea is developed by giving common
mistakes of other users as misleading hints in our work. Remolina et al. (2009) introduce
63
Automated Role Players (ARPs) in their simulation-based Intelligent Tutoring System for
training Tactical Action Officers (TAOs). ARPs also tend to make mistakes intentionally.
TAO should notice the mistake and take action. If the TAO does not respond, ARPs or ARP
Captain intervene and correct the situation themselves. This is a similar approach as
supportive agents give misleading hints. At the end, tutor agent corrects the mistake as ARP
Captain does. Learning from mistakes is one of the learning methodologies. Trying to
mislead learner with a common mistake is also used in Mengelle et al. (1998). The benefit of
learning by making mistakes is also discussed in (Bak & Chialvo, 1999; Dweck, 2007; Izzo
& Mazzone, 2008).
According to the result of the evaluation survey 56% of the participants agree with
the idea of supportive agents that give misleading hints has positive effect on their learning
process, whereas 19% of the participants reject the idea.
6.2.3 Tracking Learner’s Movements
Another methodology used in practice part was tracking learner’s movements. When
learner makes same mistakes again and again, system thinks that there would be a problem
In order to evaluate the methodology that tracks user movements, a correlation between
number of mistakes and practice score is examined. The hypothesis is that how much learner
made a mistake affects their practice score. However, the correlation coefficient (0.0002) and
Figure 5.6 show that there is no correlation between variables practice score and number of
mistakes. On the other hand, according to the practice results, it is also seen that participants
that made mistakes also get high scores from practices. It can be concluded that the
methodology of tracking users has the ability of teaching subject to the inexperienced users
and increasing their learning achievement.
According to the survey results (Figure 5.24) 47% of the participants agree with the
idea of tracking user movements and redirecting them to content materials, whereas 44% of
the participants reject the idea. The reason for the high percentage of disagreement might not
be related with the failure of the methodology, on the contrary it can be related to exhausting
effects of redirecting user from practice screen.
64
6.3 Effective Feedback
Fossati et al. (2009) develop an ITS system which helps students to understand the
subject of Linked Lists easily which is an important topic in Computer Science. It is believed
that most of the computer science students have difficulties in learning this subject.
According to the feedback of students, system is considered interesting, useful and its
performance is getting closer to the performance of human tutors. In UML-ALF intelligent
agents and computer based tutors also can give effective feedback to the learners which is
close to the performance of human tutors. The result of this proposition is determined by the
evaluation surveys which were filled out by participants after experiencing UML-ALF.
According to the results of these surveys, the feedback which is given by intelligent agent
and computer based tutors in this study is satisfactory; however it is still not so close to
human tutors.
Since it is a relative parameter that feedback is effective or not, in order to make
more accurate and quantitative decisions, in future works, different learning platforms can be
compared with a larger sample of participants.
Kenny and Pahl (2009) present an interactive training and learning environment. It
has an automated and adaptive tutoring system at the core. This system supports the
Structured Query Language (SQL) part of a database courseware system. They evaluate their
system with surveys. Forty percent agreed that their feedback mechanism is helpful whereas
feedback mechanism of UML-ALF is found 63% useful(Figure 5.20). Over twofifths of
students agreed that their guidance feature is useful. The contribution of supportive agents is
discussed in Part 6.1 and 6.2. Also in our evaluation survey 69% of the participants agreed
that supportive agents are very helpful.
65
CHAPTER 7
CONCLUSION and FUTURE WORK
Interactive and adaptive learning environments differs substantially from traditional
instructor-oriented, classroom-based teaching.This study describes an agent-based adaptive
learning framework whose goal is to implement effective tutoring system with the help of
Artificial Intelligence (AI) techniques and cognitive didactic methods into Adaptive
Educational Hypermedia Systems (AEHS). There are three main goals of this study. First
goal is to explore how supportive agents affect student’s learning achievement in distance
learning. Second goal is to examine the interaction between supportive agents and learners
with the help of experiments in Human Computer Interaction laboratories and system
analysis. The effects of the methodology that agents give misleading hints which are
common mistakes of other learners are also investigated. Last goal is to deliver effective
feedback to students both from IAs and tutors.
Learning is usually divided into two categories – knowledge acquisition and skills
training. When learners are actively engaged, they reach a higher level of understanding
while the types are different. There should be a high level of knowledge-level and skills-
level interaction between the student and a tutoring system. Scaffolding and feedback is the
most important feature of the apprenticeship theory. (Kenny & Pahl, 2009)
It is expected that this study determines how adaptive hypermedia systems and
intelligent agents contribute on distance education. Besides, this study gathers the
information of which methodologies of adaptive and intelligent techniques have positive
effects on learning, which learning path the participants follow, how they react to the
learning system, with the help of analysing data of intelligent agents, face-to-face trials and
observations of experiments in Human Computer Interaction Labs.
Considering the experience gained during this study and the findings of this study,
some recommendations for further research are provided. This study is conducted based on
66
providing an adaptive learning framework for all domains. Thus, this framework might be
used for different domains or multi-domain implementations in a future work.This study is
implemented on UML domain with semi-professional content materials and small sample
group. In a future work for better results, more professional materials – contents, exercises,
practices – and broader sample group might be used.
67
REFERENCES
Bell, A.E. (2004). Death by UML Fever, ACM Queue, 2(1), 72-80. Bowman, M., Lopez Jr., A. M., Donlon, J. & Tecuci, G. (2001). Teaching Intelligent Agents:
Software Design Methodology. CrossTalk - the Journal of the Defense Software Engineering, 14(6), 10-14, Retrieved December 27, 2009 from http://www.stsc.hill.af.mil/crossTalk/2001/06/bowman.html
Brusilovsky, P. (2000). Adaptive Hypermedia: From Intelligent Tutoring Systems to Web-
Based Education. Lecture Notes in Computer Science, Proceedings of the 5th International Conference on Intelligent Tutoring Systems, 1839, 156-169.
Brusilovsky, P. (2001). Adaptive Hypermedia. User Modeling and User-Adapted
Instruction, 11(1-2), 87-110. Brusilovsky, P. & Peylo, C. (2003). Adaptive and Intelligent Web-Based Educational
Systems. J. Artificial Intelligence in Education, 13, 156-169. Cabukovski, V.E. (2006). An Agent-Based Testing Subsystem in an E-Learning
Environment. Proceedings of the 2006 IEEE/WIC/ACM international conference on Web Intelligence and Intelligent Agent Technology, 622-625.
Chen, T. (2008). Mutual Model of Multiagent Process in E-learning Architecture. ISISE '08,
International Symposium on Information Science and Engineering, 2, 154-158. Chialvo, D. R. & Bak, P. (1999). Learning from mistakes. Neuroscience, 90(4), 1137-1148. Dweck, C.S. (2007). Perils and Promise of Praise. Educational Leadership – Early
Intervention at Every Age, 65(2), 34-39.
68
Faria, L., Silva, A., Vale, Z. & Marques, A. (2009). Training Control Centers’ Operators in
Incident Diagnosis and Power Restoration Using Intelligent Tutoring Systems. IEEE Transactions on Learning Technologies, 2(2), 135-147.
Feng, M., Heffernan, N.T., Heffernan, C. & Mani, M. (2009). Using Mixed-Effects
Modeling to Analyze Different Grain-Sized Skill Models in an Intelligent Tutoring System. IEEE Transactions on Learning Technologies, 2(2), 79-92.
Fossati, D., Eugenio, B.D., Brown, C.W., Ohlsson, S., Cosejo, D.G. & Chen, L. (2009).
Supporting Computer Science Curriculum Exploring and Learning Linked Lists with iList. IEEE Transactions on Learning Technologies, 2(2), 107-120.
Hambleton, R. K. & Swaminathan, H. (1985). Item Response Theory. Boston, MA: Kluwer-
Nijhoff. Harris, T. (2009). yUML, yUML beta v0.18. Retrieved December 27, 2009 from
http://yuml.me Hong, C.M., Chen, C.M., Chang, M.H. & Chen, S.C. (2007). Intelligent Web-based Tutoring
System with Personalized Learning Path Guidance. Seventh IEEE International Conference on Advanced Learning Technologies, 512-516.
Huffman, S.B. (1994). Instructable Autonomous Agents. PhD Thesis, University of
Michigan, Dept of Electrical Engineering and Computer Science. IMS (2001). IMS Learner Information Packaging. Retrieved June 6, 2010 from
http://www.imsproject.org/profiles/lipbest01a.html Izzo, M. F. & Mazzone, G. (2008). Learning by Doing Mistakes Improving ICT Systems
Through the Evaluation of Application Mistakes. Interdisciplinary Aspects of Information Systems Studies, Physica-Verlag HD, 123-129.
Jin, L. L. & Ming, H. C. (2008). An Intelligent Agent for Personalized E-Learning, The
Eighth International Conference on Intelligent Systems Design and Applications, 1, 27-31.
69
Keller, F. S. & Sherman, J. G. (1974). The Keller Plan Handbook. Menlo Park, CA: W. A.
Benjamin. Kenny, C. and Pahl, C. (2009). Intelligent and adaptive tutoring for active learning and
training environments, Interactive Learning Environments, 17(2), 181 — 195
Kızılkaya, G. and Aşkar, P. (2008). The effect of an pedagogical agent on the students' science achievement, Interactive Technology and Smart Education, 5(4), 208-216.
Kolb, D.A. (1984). Experiential Learning: Experience as the Source of Learning and
Development. Prentice-Hall, Inc. Kolen, M. J., Hanson, B. A. & Brennan, R. L. (1992). Conditional standard errors of
measurement for scale scores. J. Educ. Measur., 29, 285–307. Lynch, G., Palmiter, S. & Tilt, C. (1999). The Max Model: A Standard Web Site User
Model. 5th Conference on Human Factors & the Web Mengelle, T., de Léan, C. & Frasson, C. (1998). Teaching and Learning with Intelligent
Agents: Actors. Proceedings of the 4th International Conference on Intelligent Tutoring Systems, Heidelberg, Germany: Springer-Verlag, 1452, 284 – 293,
Papadimitriou, A., Grigoriadou, M. & Gyftodimos, G. (2009). Interactive Problem Solving
Support in the Adaptive Educational Hypermedia System MATHEMA. IEEE Transactions on Learning Technologies, 2(2), 93-106.
Patton, M.Q. (2002). Qualitative Research and Evaluation Methods (3rd ed.). Thousand
Oaks: Sage Publications Remolina, E., Ramachandran, S., Stottler, R. & Davis, A. (2009). Rehearsing Naval Tactical
Situations Using Simulated Teammates and an Automated Tutor. IEEE Transactions on Learning Technologies, 2(2), 148-156.
70
Russell, S. & Norvig, P. (2003). Artificial Intelligence: A Modern Approach (AIMA) (2nd ed.). Prentice Hall, New Jersey.
Scott, W. L., Beichner, R. J., Titus, A. & Martin, L. (2000). Education research using Web-
based assessment systems. J. Res. Comput. Educ., 33, 28–43.
Shih, T. K., Lin, N. H., Change, H. (2003). An intelligent e-learning system with authoring
and assessment mechanism. IEEE Int. Conf. Advanced Information Networking Applications, 782–787.
Somyürek, S. (2009). Adaptive Learning Environments: A New Paradigm in Educational
Hypermedia Design. Bilişim Teknolojileri Dergisi, 2(1), 29-38. Tecuci, G. & Hieb, M.R. (1996). Teaching intelligent Agents: the Disciple Approach.
International Journal of Human-Computer Interaction, 8(3), 259-285, 1996. Uhr, L. (1969). Teaching machine programs that generate problems as a function of
interaction with students. ACM Annual Conference/Annual Meeting Proceedings of the 1969 24th national conference, 125 – 134.
71
APPENDIX A
SCENARIOS of BUILDING RESOURCES
[U] User, [DA] Diagnostic Agent, [LA] Language Agent, [FM] Feedback Mechanism, [SA]
Supportive Agent
[SA1], [SA2] are helping agents whereas [SA3], [SA4] are misleading ones.
[SA2], [SA4] are forced to response.
[DA]: - checkElement(38)
Rule ID
Rule Syntax Practice ID Diagram Type Element Type
Rule Owner
38 [Company]+1->*[Employee]
1 1(class-diagram)
5 (relations) 7(student)
- generateReaction(38)
Rule ID
Rule Syntax Practice ID Diagram Type Element Type
Rule Owner
39 [Company]-0..*>[Employee]
1 1(class-diagram)
5 (relations) 1(DA)
40 (Company)<(Employee) 1 2(use-case dia.)
5 (relations) 1(DA)
41 The relationship between those subjects is not ‘Composition’.
1 - 9(general hint)
1(DA)
- send to SAs
[SA1] [SA2] [SA3] [SA4] - evaluateReactions() - 39:unknown - 40:false - 41:false - selects no response
- evaluateReactions() - 39:unknown - 40:false - 41:false - selects 39
- evaluateReactions() - 39:true - 40:false - 41:unknown - selects 40
- evaluateReactions() - 39:true - 40:true - 41:unknown - selects 41
Scenarios
[U] User, [DA] Diagnostic Agent, [LA] Language Agent, [FM] Feedback Mechanism, [SA]
Supportive Agent
72
Scenario 1
[S]: - Selects UML diagram type (e.g. Class-Diagram)
- Selects element type (e.g. Class)
- Enters name (e.g. Employee)
- sends information to LA
[LA]: - creates rule package
Rule ID
Rule Syntax Practice ID
Diagram Type
Element Type Rule Owner
25 [Employee] 1 1 (class-diagram)
3 (class) 7 (student)
- send package to DA
[DA]: - checkElement(25) ---> OK!
- generateReaction(25)
- send to SAs
[SAs]: - decideReaction()
Scenario 2
[S]: - Selects UML diagram type (e.g. Class-Diagram)
- Selects element type (e.g. Class)
- Enters name (e.g. User)
- sends information to LA
[LA]: - creates rule package
Rule
ID
Rule
Syntax
Practice
ID
Diagram Type Element Type Rule
Owner
26 [User] 1 1 (class-diagram) 3 (class) 7 (student)
- send package to DA
[DA]: - checkElement(26) ---> Fail!
- generateReaction(26)
- send to SAs
[SAs]: - decideReaction()
Scenario 3
[S]: - Asks help from SA2
[SA2]: - getReactions(1) (sending practice ID)
[DA]: - loadPractice(1)
73
- generateReaction(1)
- send to SA2
[SAs]: - decideReaction()
Scenario 4
[S]: - finishPractice()
[DA]: - checkPractice(1)
---> OK!
[DA]: - send results to FM
[FM]: - response positive feedback
---> Fail!
[DA]: - diagnosePractice(1)
[DA]: - send results to FM
[FM]: - response intelligent negative feedback
74
APPENDIX B
SURVEY FOR DETERMINING EXPERIMENT GROUPS
1. Uzmanlık Alanınız : ………………………………………
Bilgisayar Bilimleri
Bilgi Teknolojileri
Eğitim
Diğer
2. Unified Modeling Language (UML) ile ilgili ders aldınız mı?
Evet
Hayır
3. Unified Modeling Language (UML) iş yaşamınızda kullanıyor musunuz?
Evet
Hayır
4. Unified Modeling Language (UML) bilgi seviyenizi nasıl tanımlarsınız?
Hiç
Az
Orta
İleri
5. Unified Modeling Language (UML) ile ne kadar ilgileniyorsunuz?
Hiç
Az
Orta
Çok
75
APPENDIX C
EVALUATION SURVEY
Çoktan Seçmeli Sorular
1. Eğitimlerin içeriği bilgi bakımından yeterliydi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
2. Eğitimlerin içeriği görsellik bakımından yeterliydi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
3. Eğitimleri hangi sırada almam gerektiği açıktı.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
4. Eğitimlerin öğrenmemde katkısı olduğunu düşünüyorum.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
76
5. Bana sunulan eğitim yolu benim için uygundu.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
6. Eğitimler, eğitim sonrasında karşıma çıkan testi yapmamda faydalıydı.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
7. Eğitimlerin içeriğinde zaten hakim olduğum konular bulunuyordu.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
8. Eğitimler gereksiz ve fazla bilgi içeriyordu.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
9. Eğitimleri kendi tercih ettiğim sırada almak isterdim.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
10. Eğitim ekranlarında kullanıcıyı yönlendirme yetersizdi.
o Kesinlikle katılmıyorum
o Katılmıyorum
77
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
11. Eğitim ekranları kullanılabilirlik açısından yetersizdi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
12. Alıştırma kısmında ajanların yardımları çok faydalıydı.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
13. Alıştırma kısmında ajanların yorumları beni olumsuz etkiledi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
14. Alıştırma ekranlarındaki öğretmenin geri bildirimleri, gerçek öğretmenin geri bildirimlerine yakındı.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
15. Alıştırma ekranları bilgimi sınamama ve geliştirmeme yardımcı oldu.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
78
16. Alıştırma ekranları kullanılabilirlik açısından yetersizdi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
17. Alıştırma ekranlarındaki geri bildirimler yeterliydi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
18. Alıştırma ekranlarındaki yönlendirmelerin motivasyonumu arttırmada etkiliydi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
19. Alıştırmalar, eğitimlerde öğrendiğim bilgiler ile alakalı değildi.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
20. Alıştırma ekranlarında bazı ajanların hatalı cevaplarının konuyu öğrenmemde olumlu etkisi oldu.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
21. Konuyu öğrenmemde, alıştırma ekranları eğitim ekranlarından daha çok yardımcı oldu.
o Kesinlikle katılmıyorum
o Katılmıyorum
79
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
22. Sistem genel olarak UML konusundaki bilgimi arttırmada başarılıydı.
o Kesinlikle katılmıyorum
o Katılmıyorum
o Fikrim yok
o Katılıyorum
o Kesinlikle katılıyorum
80
APPENDIX D
HCI LAB RESULTS
Figure 8.1 HCI participant 1 Class Diagram Practice
Figure 8.2 HCI participant 1 Use Case Diagram Practice
81
Figure 8.3 HCI participant 1 Activity Diagram Practice
Figure 8.4 HCI participant 2 Class Diagram Practice
Figure 8.5 HCI participant 2 Use Case Diagram Practice
82
Figure 8.6 HCI participant 2 Activity Diagram Practice
Figure 8.7 HCI participant 3 Class Diagram Practice
Figure 8.8 HCI participant 3 Use Case Diagram Practice
83
Figure 8.9 HCI participant 3 Activity Diagram Practice
Figure 8.10 HCI participant 4 Class Diagram Practice
Figure 8.11 HCI participant 5 Class Diagram Practice
84
APPENDIX E
APPROVAL FROM ETHICS COMMITTEE
85
APPENDIX F
TEST QUESTIONS
CHAPTER 1 INTRODUCTION
1. Which is NOT true about UML?
• UML is a visual language for modeling and communicating about systems through the use of diagrams and supporting text.
• UML is a programming language. • UML brings together the information systems and technology industry's best
engineering practices. • UML stands for Unified Modeling Language.
2. Which is NOT an aspect of UML?
• Language • Methodology • Unified • Modeling
3. Which is a false statement about model?
• A model is a representation of a subject. • A model captures a set of ideas known as abstractions about its subject. • It provides a common understanding of the requirements and the system for team
members. • When creating a model, everything should be represented about the subject all at
once. 4. The UML is a language for ... [2 answers]
• Specifying • Manufacturing • Constructing • Programming
86
5. Which is not one of the methods of OMG's scope when creating the UML?
• Yourdon-Coad • Booch Method • Object-modeling technique (OMT) • Object-oriented software engineering (OOSE)
6. Who are the 'Three Amigos'?
• Edward Yourdon - Grady Booch - Ward Cunningham • James Rumbaugh - Grady Booch - Ivar Jacobson • Ward Cunningham - Kent Beck - James Rumbaugh • Ivar Jacobson - Kent Beck - Edward Yourdon
7. Which of the following lifecycle is in correct order?
• Analysis - Design - Deployment – Testing • Design - Requirement Gathering - Implementation – Deployment • Requirement Gathering - Analysis - Implementation – Testing • Analysis - Design - Testing – Implementation
8. Which of the following statement is false?
• Within a waterfall approach, all the elements are captured and analyzed, and the whole system is designed, implemented, tested, and deployed in this linear sequence.
• The UML models must be fairly complete at each step. • When applying an iterative approach, any subsets of the lifecycle activities are
performed several times to better understand the requirements and gradually develop a more robust system.
• When applying an iterative approach, solving problems discovered during testing activities might be too late or too costly.
9. Which is NOT one of the benefits of an iterative approach?
• We can better manage complexity by building a system in smaller increments rather than all at once.
• We can better manage changing requirements by incorporating changes throughout the process and not trying to capture and address all the requirements at once.
• We can provide general solutions to users at the end of the process. • We can solicit feedback from users concerning the parts of the system already
developed, so we may make changes and guide our progress in providing a more robust system that meets their requirements.
10. Which of the following statement is false?
• A use case is a functional requirement described from the perspective of the users of a system.
• The elements and how they interact and collaborate is known as the system's structure. Modeling a system's structure is known as structural modeling.
87
• An architecture-centric process focuses on the architecture of a system across iterations.
• A risk is any obstacle or unknown that may hinder our success.
CHAPTER 2 OBJECT ORIENTED MODELING
1. Which of the following statements are true? [2 answers]
• A project uses requirements as an input and produces a system (or a part of a system) as an output.
• Database of the project management system executes on a client computer. • When creating a project, start and end dates of the project is uncertain. • Project becomes active when human resources are assigned to the project.
2. Which of the following statement is false?
• An alphabet defines the simplest parts of a language: letters, characters, signs, and marks.
• The smallest units of meaning in a language are its "letters". • A language's syntax specifies the notation used for communication and is determined
by the language's alphabet. 3. How can we simply describe the classes?
• The general concepts shown in the sentences • The specific concepts • The general relationships • The specific relationships
4. Which of the facts below about the attributes IS NOT correct?
• Something that an object knows, which essentially represents data, is called an attribute.
• A class defines attributes and an object has values for those attributes. • If two objects have the same values for their attributes, then they are counted as
same object. • Associations and attributes are known as structural features.
5. Which of the following facts about messages and stimuli IS NOT correct?
• A stimulus is an instance of a message similar to how an abject is an instance of a class.
• The sender is known as the client. • The receiver is known as the supplier. • Communication between objects via their links is called a message.
6. Encapsulation is also known as...
• Localization • Information hiding • Abstraction • Generalization
88
7. The ability to have multiple methods for a single operation is called... • Encapsulation • Generalization • Polymorphism • Abstraction
8. Which of the following diagrams DOES NOT belong to Structural Modeling?
• Use-case diagrams • Sequence diagrams • Deployment diagrams • Class diagrams
9. Which of the following IS one of the types of elements of Activity Diagrams?
• An event • Associations • A lifeline • A swimlane
10. The UML diagrams may be generally organized around:
• The use-case or user architectural view • The structural or static architectural view • The behavioral or dynamic architectural view • All of the above
CHAPTER 3 CLASS AND OBJECT DIAGRAMS
1. Which of the following statement is false? • Class modeling is a specialized type of modeling concerned with the general
structure of a system. • Object modeling is a specialized type of modeling concerned with the structure of a
system at a particular point in time. • Object modeling is usually used in conjunction with class modeling to explore and
refine class diagrams. • Class and object modeling usually start before the requirement analysis.
2. Which of the following statements are true? [2 answers]
• Structural features define what objects of the class know. • Behavioral features define what objects of the class can do. • Structural features include operations and methods. • Behavioral features include attributes and associations.
3. Which of the following statement is true?
• An attribute is what an object of a class can do. It is a specification of a service provided by the object.
• An operation is what an object of a class knows. It's an element of data maintained by the object.
• In a class diagram, attributes are listed in the second compartment for a class. • In a class diagram, operations are listed in the first compartment for a class.
89
4. Which of the following fact about objects is false? • An object is a general concept. It defines a type of class and its characteristics. • An object has structural features including attribute values and links. • An object has behavioral features including operations and methods. • An object is shown as a solid-outline rectangle with two standard compartments
separated by horizontal lines. 5. Which of the following fact about associations is false?
• An association defines a type of link and is a general relationship between classes. • A binary association relates two classes. • A binary association may be labeled with a name. • A binary association should be named using a noun phrase.
6. Which of the following facts about multiplicity is true? [2 answers]
• It indicates how many objects of a class may relate to the other classes in an association.
• Default multiplicity for association end is zero or more. • Multiplicity is a must. • Multiplicity is simply undefined, unless it is specified.
7. Which of the following statement about aggregation is false?
• Aggregation is whole-part relationship between an aggregate, the whole, and its parts.
• Aggregation is often known as a has-a relationship. • Aggregation is shown using a filled-in diamond attached to the class that represents
the whole. • Aggregation is used only with binary associations.
8. Which of the following statement about composition is false?
• Composition is whole-part relationship between a composite and its parts. • Composition is often known as a has-a relationship. • Composition is shown using a filled-in diamond attached to the class that represents
the whole. • Composition is used only with binary associations.
9. Which is not a rule for representing links?
• Label links with their association names, and underline the names to show that they are specific instances of their respective associations.
• Ensure that link ends are consistent with their corresponding association ends. • Translate association multiplicity into one or more specific links between specific
objects. • A link relates two classes.
CHAPTER 4 USE CASE DIAGRAMS
1. Which of the following fact about use cases is false? • Given that every project has limited resources, you can use this information about
use cases to determine how best to execute a project.
90
• Use cases cannot be reused. • Use-case modeling is a specialized type of structural modeling concerned with
modeling the functionality of a system. • An understanding of how use cases are related allows users and analysts to negotiate
and reach agreement concerning the scope and requirements of a system.
2. Which is not an element of a use case diagram? • Use Case • Activity • Actor • Extends
3. Which element is shown with the specified figure?
• Use Case • Admin • Actor • User
4. Which of the following statements are true? [2 Answers] • An actor can be shown as a class marked with the actor keyword and labeled with
the name of the actor class. • An actor may be a human user or another system. • An actor does not represent a resource owned by another project or purchased rather
than built. • When speaking of a specific actor of a class, the term actor class is used.
5. Which of the following statement is false?
• Use case is shown as an ellipse and labeled with the name of the use-case class. • Use cases may be enclosed by a rectangle that represents the boundary of the system
that provides the functionality. • A use case is a functional requirement that is described from the perspective of the
users of a system. • When speaking of a class of use cases, it's customary to use the term use-case
instance. 6. What is communicate associations?
• It is a functional requirement that is described from the perspective of the users of a system.
• It indicates how many use-case classes may relate to the other classes in an association.
• It addresses the question of how actors and use cases are related and which actors participate in or initiate use cases.
• It is the performance of a specific sequence of actions and interactions. 7. Which of the following fact about include dependency is false?
• An include dependency from one use case to another use case indicates that the base use case will include or call the inclusion use case.
91
• An include dependency is shown as a dashed arrow from the inclusion use case to the base use case marked with the include keyword.
• The base use case is responsible for identifying where in its behavior sequence or at which step to include the inclusion use case.
• An include dependency can be used when a use case may be common to multiple other use cases and is therefore factored out of the different use cases so that it may be reused.
8. Which of the following fact about extend dependency is false?
• An extend dependency can be used to address the situation by factoring out optional behavior from a use case.
• A use case may extend multiple use cases, and a use case may be extended by multiple use cases.
• An extend dependency from one use case to another use case indicates that the extension use case will extend and augment the base use case.
• An extend dependency is shown as a solid arrow from the extension use case to the base use case marked with the extend keyword.
CHAPTER 5 COMPONENT AND DEPLOYMENT DIAGRAMS 1. What is the difference between Component modeling and Deployment modeling?
• Component modeling is structural, Deployment modeling is behavioral. • Deployment model shows the external resources that those components require. • Component modeling typically starts after the design of the system is fairly
complete. • Deployment modeling is applied during design activities.
2. Which of the following can be a component of a system?
• A user interface component. • A business-processing component. • A data component. • All of the above.
3. Which of the following facts about components is false?
• A component is shown as a rectangle. The rectangle is labeled with the name of the component class fully underlined.
• A component is shown as a rectangle with two small rectangles protruding from its side.
• A component instance is shown similar to a component class, but is labeled with the component instance name followed by a colon followed by the component class name.
• Both names are optional, and the colon is present only if the component class name is specified.
4. Which of the following facts about nodes is false?
• A node is a resource that is available during execution time. • A node may be a computer, printer, server, Internet, or any other kind of resource
available to components.
92
• A node is shown as an ellipse labeled with the node's name. • When speaking of a specific component of a class, the term node instance is used.
5. Which of the following statement is false?
• A reside dependency from a component to any UML element indicates that the component is a client of the element, which is itself considered a supplier, and that the element resides in the component.
• A use dependency from a client component to a supplier component indicates that the client component uses or depends on the supplier component.
• A use dependency from a client component to a supplier component's interface indicates that the client component uses or depends on the interface provided by the supplier component.
• A deploy dependency from a client component to a supplier node indicates that the supplier component is deployed on the client node.
CHAPTER 6 SEQUENCE AND COLLABORATION DIAGRAMS 1. Which of the following fact about sequence diagrams is false?
• A sequence diagram shows elements as they interact over time, showing an interaction or interaction instance.
• Sequence diagrams are organized along two axes: the horizontal axis shows the elements that are involved in the interaction, and the vertical axis represents time proceeding down the page.
• The elements on the vertical axis may appear in any order. • Sequence diagrams are made up of a number of elements, including class roles,
specific objects, lifelines, and activations. 2. Which of the following elements are the elements of sequence diagram? [2 answers]
• Class roles • Association roles • Lifelines • Specific links
3. Which of the following element is not an element of collaboration diagram?
• Class roles • Specific objects • Specific links • Activation
4. Which element is shown with the specified figure?
• Class roles • Specific objects • Lifelines • Activation
5. Which of the following statement is false?
• Reflexive communication is an element that communicates with itself where a communication is sent from the element to itself.
93
• When an element is created during an interaction, the communication that creates the element is shown with its arrowhead to the element.
• When an element is destroyed during an interaction, the communication that destroys the element is sent back from the element to communication's lifeline where the destruction is marked with a large "X" symbol.
• Repetition (which involves repeating a set of messages or stimuli) within a generic-form interaction is shown as a set of communications enclosed in a rectangle.
6. Which is not a difference between collaboration and sequential diagram?
• A collaboration diagram shows elements as they interact over time. • Collaboration diagrams are time and space-oriented and emphasize the overall
interaction, the elements involved, and their relationships. • Sequence diagrams are useful for complex interactions. • Collaboration diagrams are useful for visualizing the impact of an interaction on the
various elements.
CHAPTER 7 STATE DIAGRAMS 1. Which of the following statement is false?
• Each element has a lifecycle. • A state is a specific condition or situation of an element during its lifecycle. • The current state of an element is called its active state. • A state diagram may have only one initial and final state.
2. Which of the following statement is false?
• An initial state indicates the state of an element when it is created and it is shown using a small solid filled circle.
• A final state indicates the state of an element when it is destroyed and it is shown using a circle surrounding a small solid empty circle.
• There are various types of states, including simple, initial, and final states. • State modeling is a specialized type of behavioral modeling concerned with
modeling the lifecycle of an element. 3. Which is not a simple state?
• Inactive • Active • Suspended • Terminated
4. Which of the following facts about transitions is false?
• A transition without an event and action is known as an automatic transition. • When an event occurs, the transition is said to fire. • A transition is shown as a dashed line from a source state to a target state labeled
with the event followed by a forward slash followed by the action. • Transitions between states occur as follows: An element is in a source state. Then an
event occurs. After that an action is performed. Finally, the element enters a target state.
94
5. Which of the following statement is false? • An event is an occurrence, including the reception of a request. • An action represents processing. • A state may be shown as a circle. • It is allowed to put one or more state diagrams inside a single state to indicate that
when an element is in that state, other elements inside of it have their own states. CHAPTER 8 ACTIVITY DIAGRAMS 1. Which of the following statement is false?
• Each element has the responsibility of appropriately reacting to the communications it receives.
• An action state represents processing as an element fulfills a responsibility. • A simple action state represents processing. • An activity diagram may have more than one initial and final state.
2. Which of the following statement is false?
• An initial action state indicates the first action state on an activity diagram. It is shown using a small solid empty circle.
• A final action state indicates the last action state on an activity diagram and it is shown using a circle surrounding a small solid filled circle (a bull's eye).
• There are various types of action states, including simple, initial, and final action states.
• Activity modeling is a specialized type of behavioral modeling concerned with modeling the activities and responsibilities of elements.
3. Which is not an activity diagram element?
• Initial Action State • Final Action State • Class • Decision
4. Which of the following facts about transitions is false?
• A flow transition shows how action states are ordered or sequenced. • A control-flow transition indicates the order of action states. • An object-flow transition indicates that an action state inputs or outputs an object. • Object-flow transitions are also known as default transitions or automatic transitions.
5. Which of the following facts about transitions is false?
• A swimlane is a visual region in an activity diagram that indicates the element that has responsibility for action states within the region.
• A decision involves selecting one control-flow transition out of many control-flow transitions based upon a condition.
• Concurrency involves selecting multiple transitions simultaneously. • If a bar has one incoming transition and two or more outgoing transitions, it
indicates that all outgoing transitions occur once the incoming transition occurs. This is called synchronization of control.
95
top related